# Agent Skills
Source: https://docs.modularmind.app/capabilities/agent-skills
Extend Maia with specialized knowledge, workflows, and domain expertise using the Agent Skills standard
## Give Maia New Capabilities
Agent Skills allow you to package specialized knowledge, workflows, and domain expertise into reusable capabilities that Maia can discover and use automatically. Whether you need brand guidelines enforcement, custom analysis pipelines, or company-specific procedures, skills give Maia the context it needs to work reliably.
## What are Agent Skills?
Agent Skills are a simple, open format for extending AI agents with specialized capabilities. A skill is a folder containing instructions, scripts, and resources that agents can load on demand when relevant to a task.
The format is built on three principles:
Skills are human-readable folders with markdown instructions
Use the same skill across different AI agents and platforms
Load only what's needed, keeping context efficient
### The Agent Skills Standard
Each skill contains a `SKILL.md` file with:
* **Metadata**: Name and description (so Maia knows when to use it)
* **Instructions**: Step-by-step guidance in markdown format
* **Optional Resources**: Scripts, templates, reference documents, and assets
This open format means skills you create for Maia can work with other skills-compatible agents like Claude AI, Claude Code, Gemini CLI, and OpenAI Codex.
## How Maia Uses Agent Skills
Maia integrates Agent Skills using a progressive disclosure approach:
1. **Discovery**: At startup, Maia loads the name and description of all your skills
2. **Activation**: When a task matches a skill's description, Maia loads the full instructions
3. **Execution**: Maia follows the instructions, accessing scripts and resources as needed
This keeps Maia fast while giving it access to extensive specialized knowledge on demand.
### Automatic Integration
Unlike other AI agents that require manual skill installation, **all skills in Maia are immediately available across**:
* ✓ All conversation sessions
* ✓ All workflows (including existing ones)
* ✓ All deployed workflows
* ✓ All workflow triggers (routine, webhook, and action)
No configuration, no restarts, no installation steps. Create a skill and Maia can use it instantly.
## Managing Skills in Maia
Access skill management through **Skills** settings where you can create, upload, and manage all your agent skills.
### Creating New Skills
Maia can generate complete Agent Skills for you from any source material:
Navigate to Skills settings in your Maia workspace from the menu at the bottom left. Click the **Create Skill** button and select **Let Maia Create**.
Tell Maia what capability you want to create:
```
Create a brand guidelines skill that enforces:
- Logo usage rules (primary logo for white backgrounds)
- Color palette: Primary #39DA5D, Secondary #1A1A1A
- Typography: Inter for headings, SF Pro for body text
- Never use gradients or shadows in official materials
```
Optionally attach PDFs, text files, or documents containing:
* Existing guidelines or procedures
* Example workflows
* Technical specifications
* Company policies
Maia generates a complete skill with proper structure. Review the files in the built-in editor and make any adjustments.
The skill is immediately available to Maia in all contexts
### Uploading Existing Skills
If you already have Agent Skills created for other platforms or shared by teammates:
1. **Navigate to Skills settings** from the menu at the bottom left
2. **Click Create Skill** and select **Upload Skill (ZIP)**
3. **Select your ZIP file** (ensure the skill is properly packaged)
4. **Enable the skill** to make it available to Maia
### Managing Your Skills
In the Skills settings panel, you can:
Control which skills are active without deleting them
Refine skill instructions and update procedures
Remove skills you no longer need
## Use Cases
Agent Skills excel at capturing specialized knowledge and repeatable workflows:
**Ensure consistency** across all generated content
* Logo usage rules and asset selection
* Color palettes and design systems
* Typography and formatting standards
* Tone of voice and messaging guidelines
*Example: "Generate a press release following our brand guidelines" — Maia automatically applies your style rules*
**Standardize analysis workflows**
* Custom data cleaning procedures
* Industry-specific calculations
* Compliance-required transformations
* Multi-step analysis pipelines
*Example: "Analyze this sales data" — Maia follows your company's specific analysis methodology*
**Generate consistent documents**
* Contract templates and legal formats
* Report structures and sections
* Proposal frameworks
* Meeting note templates
*Example: "Create a project proposal" — Maia uses your organizational template and requirements*
**Follow coding standards**
* Code review checklists
* Testing procedures
* Deployment workflows
* Architecture patterns
*Example: "Review this code" — Maia applies your team's specific standards and practices*
**Maintain regulatory compliance**
* Approval workflows
* Document retention policies
* Privacy requirements
* Industry regulations (HIPAA, GDPR, SOC2)
*Example: "Draft a privacy notice" — Maia ensures compliance with your specific requirements*
**Package specialized knowledge**
* Industry terminology and context
* Technical specifications
* Best practices and procedures
* Historical context and decisions
*Example: "Explain this medical term" — Maia uses your healthcare organization's internal knowledge base*
## Example: Creating a Brand Guidelines Skill
Here's how Maia structures a complete Agent Skills package:
```
brand-guidelines/
├── SKILL.md # Main instructions with metadata
├── references/
│ ├── colors.md # Detailed color specifications
│ ├── typography.md # Font usage rules
│ └── examples.md # Good and bad examples
└── assets/
├── logo-primary.svg # Logo files
└── logo-white.svg
```
The `SKILL.md` file contains:
```markdown theme={null}
---
name: brand-guidelines
description: Enforces company brand guidelines including logo usage, color palette, typography, and visual style rules. Use when generating any marketing materials, presentations, or public-facing content.
---
# Brand Guidelines
## When to use this skill
Use this skill whenever creating or reviewing:
- Marketing materials
- Social media content
- Presentations
- Website copy
- Email campaigns
## Logo Usage
Always use the primary logo (#39DA5D) on white backgrounds.
Use the white logo on dark backgrounds (#1A1A1A or darker).
Never:
- Stretch or distort the logo
- Add effects (shadows, gradients, outlines)
- Place on busy backgrounds
## Color Palette
Primary: #39DA5D (Maia Green)
Secondary: #1A1A1A (Charcoal)
...
```
When you ask Maia to "create a social media post," it automatically:
1. Recognizes the task matches the brand guidelines skill
2. Loads the full instructions
3. Applies your specific brand rules
4. Generates compliant content
**Quick Start**: Try creating a skill for your brand's tone of voice. Include 3-5 key principles and some example phrases to use and avoid. Within seconds, Maia will apply this guidance to all content generation.
Learn how Maia connects to Google Workspace and other tools
# Data Analysis
Source: https://docs.modularmind.app/capabilities/data-analysis
Real-time code generation and execution for calculations, data analysis, and visualizations
## Execute Code On Demand
Maia can write and execute code in real-time to solve complex problems, perform calculations, and transform data — all within your workflow.
## Why Data Analysis Matters
### Trustworthy Mathematical Operations
When precision is critical, Maia generates and executes code to guarantee accurate results:
Financial modeling, statistical analysis, scientific computations
Clean, filter, aggregate, and restructure datasets
Run formulas against real data to verify accuracy
Process conditional logic and multi-step algorithms
**Example:**
```
Calculate the compound annual growth rate (CAGR) for a portfolio
that started at $100,000 and grew to $175,000 over 5 years.
```
Instead of estimating, Maia generates and executes Python code to return the precise result: **11.84%**.
## Data Visualization
Maia creates professional visualizations from any data source — whether you provide it directly or Maia retrieves it from the web or connected platforms.
### Visualization Types
* **Line Charts** - Trends over time
* **Bar Charts** - Comparisons across categories
* **Pie Charts** - Distribution percentages
* **Scatter Plots** - Correlations and patterns
* **Heat Maps** - Density and intensity mapping
* **Histograms** - Distribution analysis
* **Box Plots** - Statistical summaries
* Formatted data tables
* Summary statistics
* Pivot tables
* Network diagrams
* Gantt charts
* Geographic maps
* Contour plots
### From Data to Insight
Upload files, paste datasets, or have Maia gather information from APIs, databases, or web sources
```
Analyze this sales data and create:
1. A line chart showing monthly revenue trends
2. A bar chart comparing performance by region
3. A table of top 10 products by revenue
```
Maia generates code to process your data and produces publication-ready visualizations
## Your Data Analysis Teammate
Data analysis transforms Maia into a powerful analytical partner:
✅ **Process large datasets** - Handle thousands of rows efficiently\
✅ **Perform statistical analysis** - Correlations, regressions, hypothesis testing\
✅ **Generate insights** - Identify trends, outliers, and patterns\
✅ **Create reports** - Combine analysis with visualizations in documents\
✅ **Automate calculations** - Build workflows that run complex math operations
**Request:**
```
I have Q4 sales data in this CSV. Analyze it and show me:
- Total revenue by month
- Top 5 products by units sold
- Average deal size by sales rep
- Month-over-month growth rate
Create visualizations for each metric.
```
**What Maia Does:**
1. Reads and validates the CSV data
2. Executes Python code to calculate each metric
3. Generates four separate chart visualizations (line chart for revenue, bar chart for products, etc.)
4. Creates a summary table with key statistics
5. Provides numerical summaries alongside charts
**Request:**
```
Calculate the break-even point for a SaaS product with:
- Monthly fixed costs: $50,000
- Variable cost per customer: $5/month
- Subscription price: $99/month
Show me break-even in customers and create a graph showing
profitability from 0 to 1,000 customers.
```
**What Maia Does:**
1. Generates a financial model in code
2. Calculates break-even: **532 customers**
3. Creates a line graph showing the profitability curve
4. Highlights the break-even point on the visualization
**Request:**
```
Get the last 30 days of Bitcoin prices, calculate the daily
percentage change, and create a chart showing the price
movement with volume bars below.
```
**What Maia Does:**
1. Retrieves historical Bitcoin data from a financial API
2. Calculates daily price changes programmatically
3. Generates a professional price chart with OHLC data
4. Adds volume bars as a subplot
5. Returns insights about volatility and trends
## How It Works
Maia's data analysis environment:
* **Secure sandbox** - Code runs in isolated environments
* **Library access** - Popular data science libraries (pandas, matplotlib, numpy)
* **Output variety** - Return values and visualizations
## Use Cases
### Business Analytics
* Sales forecasting and trend analysis
* Customer segmentation
* KPI calculation and monitoring
* ROI and financial modeling
### Research & Science
* Statistical analysis
* Data cleaning and preparation
* Experiment result processing
* Publication-ready visualizations
### Operations
* Inventory optimization
* Resource allocation calculations
* Scheduling algorithms
* Performance benchmarking
### Reporting
* Automated dashboard generation
* Executive summaries with charts
* Comparative analysis reports
* Real-time data monitoring
Create high-fidelity visuals programmatically within your workflows
# Document Generation
Source: https://docs.modularmind.app/capabilities/document-generation
Generate and edit professional documents, spreadsheets, and presentations natively
## Production-Ready Office Files
Maia can generate and edit Word, Excel, and PowerPoint files natively — full, rich, production-ready documents without any copy-paste or manual assembly. Whether you need a 2-page summary or a 50-page report, Maia handles document creation with consistent quality and formatting.
Maia generates dozens of pages of structured content and can also **edit existing documents** you already have — not just create new ones from scratch.
## Supported Document Types
`.docx` (Microsoft Word)
**Use for:** Reports, proposals, contracts, letters, business documents
`.xlsx` (Microsoft Excel), `.csv`
**Use for:** Financial models, dashboards, data workbooks, structured datasets
`.pptx` (Microsoft PowerPoint)
**Use for:** Pitch decks, investor presentations, training materials, slide content
Standard PDFs and styled presentations
**Use for:** Final deliverables, professional reports, formatted output
`.txt`, `.md`, `.csv`
**Use for:** Data exports, simple lists, markdown content
## Editing Existing Documents
Maia doesn't just create documents from scratch — it can open, update, and refine Office files you already have. Upload a DOCX, XLSX, or PPTX file and ask Maia to:
* Update figures or data in a spreadsheet
* Revise sections of a report
* Add new slides to an existing presentation
* Reformat or restructure content
## Bring Your Own Templates
You can feed Maia your existing DOCX, XLSX, or PPTX templates and have it fill them in. Your branding, your structure, your formatting — Maia produces documents that look like they came from your team.
This is especially useful for anything that needs to stay consistent:
* Client-facing deliverables
* Weekly or monthly reports
* Pitch decks for different verticals
* HR documents and onboarding materials
* Financial statements and invoices
Upload your template as an [Executive File](/capabilities/executive-files) or attach it directly in a conversation, and Maia will use it as the foundation for generated output.
Learn how Maia builds interactive web applications
# Knowledge Base
Source: https://docs.modularmind.app/capabilities/executive-files
Give your executives a personal knowledge base with uploaded documents
## A Personal Knowledge Base
Every great executive knows their business inside out. With Executive Files, you can upload documents directly to your AI executives as a personal knowledge base. Give it once, and your executive always has access to the information it needs.
## Supported File Types
Research reports, business plans, SOPs, policies
Word documents, Excel spreadsheets, PowerPoint presentations
Plain text, markdown, CSV data, configuration files
Logos, design references, visual assets for use in generated content
## What You Can Upload
SOPs, policies, and process documentation that your executives should follow when operating
Research reports, business plans, competitive analysis, and reference materials for informed decision-making
Product specs, pricing sheets, onboarding materials, and client-facing templates
Logos, design references, and visual guidelines for use in generated images and documents
## Intelligent Context Management
Maia doesn't dump all uploaded files into every interaction. It uses intelligent context management — knowing **when** a file is relevant and pulling from it efficiently. Your documents are always available without cluttering conversations or workflows where they aren't needed.
Your executives apply this knowledge, cross-reference it, and use it to power every plan, workflow, and action they run — exactly like well-briefed human operators would.
Learn how Maia researches the live web in real time
# Human in the Loop
Source: https://docs.modularmind.app/capabilities/human-in-the-loop
Stay in control with intelligent checkpoints during complex operations
## Control When It Matters
Maia and your custom executives can pause mid-operation to ask for clarifications or request your review before proceeding with critical steps. This works for both quick one-off tasks and complex multi-step automations.
You're in the loop at every decision that matters — without being interrupted on every step.
Human in the Loop is designed to be frictionless: checkpoints only where they matter, not on every action.
## How It Works
Your executives intelligently identify moments that require human judgment and pause to check in. This happens naturally within the conversation or workflow — no special configuration required.
During execution, the executive recognizes a step that requires human input — an ambiguity, a critical action, or a point where preferences matter.
The executive pauses, presents what it's about to do (or what it needs clarification on), and waits for your input.
Approve, adjust, or redirect. Your executive incorporates your decision and continues.
## Real-World Examples
Before sending a bulk outreach sequence, Maia surfaces the final draft and waits for your approval before triggering delivery.
Mid-workflow, an executive flags an unclear decision — like which client account to update — and asks you directly instead of guessing.
For long-running automations, Maia checkpoints at key stages, giving you the chance to adjust course before the next phase kicks off.
Data deletions, overwrites, or irreversible actions pause for confirmation rather than proceeding automatically.
## Works Everywhere
Human in the Loop works in both interactive sessions and deployed workflows. Even during automated runs triggered by schedules or webhooks, your executives can reach out for clarification when they encounter something that needs your judgment.
Extend Maia with specialized knowledge and domain expertise
# Image Generation
Source: https://docs.modularmind.app/capabilities/image-generation
Create high-fidelity visuals programmatically within your workflows
## Native Visual Creation
Maia can natively generate high-fidelity images, allowing you to move from concept to final visual asset without ever leaving the app. Whether you need a hero image for a blog post or a specific graphic for a presentation, Maia handles the creation programmatically.
Image generation integrates directly into your workflows, enabling automated visual content creation at scale.
## Key Features
Create crisp, high-resolution visuals ready for any professional use
Choose from 16:9, 4:3, 1:1, and more to perfectly fit your destination platform
Include image generation as a step in your multi-stage automations
Generate visuals based on dynamic inputs and data from your workflows
## Use Cases
### Content Creation
* **Blog Hero Images** - Generate featured images that match your article's theme
* **Social Media Graphics** - Create platform-optimized visuals for posts and ads
* **Newsletter Headers** - Design consistent visual branding for email campaigns
### Business & Marketing
* **Presentation Graphics** - Build custom visuals for slides and reports
* **Product Mockups** - Generate concept images for product ideas
* **Marketing Collateral** - Create banners, thumbnails, and promotional imagery
### Automated Pipelines
* **Report Generation** - Add relevant visuals to automated reports
* **Content Workflows** - Generate images as part of end-to-end content creation
* **Dynamic Outputs** - Create visuals that adapt based on data inputs
## Aspect Ratio Options
* **16:9** - Ideal for presentations, YouTube thumbnails, website headers
* **4:3** - Classic format for documents and traditional displays
* **3:2** - Photography standard, great for print materials
* **9:16** - Perfect for Instagram Stories, TikTok, mobile content
* **3:4** - Vertical documents and mobile-first designs
* **2:3** - Portrait photography and vertical banners
* **1:1** - Instagram posts, profile images, icons
## Example Workflow
Describe what you need: "Create a hero image for a blog post about AI in healthcare"
Set quality, aspect ratio, and style preferences for your output
Maia creates your image based on the description and parameters
Use the generated image in documents, websites, or pass it to other workflow steps
Package specialized capabilities and domain expertise into reusable skills
# Memory
Source: https://docs.modularmind.app/capabilities/memory
Persistent, evolving memory that lets your executives know your business deeply
## Persistent Business Context
One of the biggest gaps between a great AI assistant and a true AI executive has always been this: it forgets you the moment the conversation ends. That changes with Memory.
Maia now has persistent memory — and more importantly, memory that **updates**. Maia doesn't just store a snapshot of who you are and move on. It actively keeps that understanding fresh, refining what it knows about your business, your preferences, and the way you work as things evolve.
## What Maia Remembers
Your goals, key details, and operational context across every conversation
Tone guidelines, working style, and communication preferences — so you never repeat yourself
A progressively deeper picture of your processes, priorities, and the standards you care about
What's been decided, what constraints apply, and what context carries forward
## Memory That Updates
Unlike static profile storage, Maia's memory is **living context**. As your business evolves — new priorities, new team members, new processes — Maia's understanding stays current, not stale.
This means your executives aren't starting from scratch each time. They build on what they already know, compounding their effectiveness with every interaction.
## How Memory Powers Executives
Memory acts as a context fabric across decoupled workflows. When an executive runs a weekly report, it knows what happened in last week's report. When it handles a client request, it remembers their preferences from previous interactions.
This is the foundation for executives that genuinely know your business — not just for one session, but as persistent, evolving operational partners.
Give your executives a personal knowledge base with uploaded documents
# The Secure Browser
Source: https://docs.modularmind.app/capabilities/secure-browser
How Maia interacts with websites like a human, securely and at scale
## Computer Use: Beyond Text Scraping
Maia doesn't just read websites, it can **use** them. The Secure Browser gives Maia the ability to click, scroll, type, and navigate websites exactly like you would.
This is **computer use** technology: Maia sees the screen, understands visual elements, and takes actions just like you do with your browser.
## What the Secure Browser Can Do
Buttons, links, menus, tabs, dropdowns
Text inputs, checkboxes, radio buttons, file uploads
Follow links, use breadcrumbs, handle pagination
Reveal hidden content, trigger lazy loading
Retrieve text, take screenshots, extract links
Close modals, accept cookies, dismiss notifications
## How It Works
### Visual Understanding
The browser doesn't use hidden APIs or HTML parsing. It sees the page as you do:
1. **Takes a screenshot** of the current view
2. **Analyzes the visual layout** (buttons, text, images)
3. **Identifies interactive elements** (clickable areas)
4. **Plans the next action** (what to click or type)
5. **Executes the action** and observes the result
This means Maia can interact with any website, even complex single-page applications that make traditional scraping difficult.
### Example: Applying to Y Combinator
**Your request:**
💬 *Apply to Y Combinator's next batch for my startup,
using information from our pitch deck and business plan.*
**What Maia does:**
Opens YC's application portal and starts a new application
Identifies all form fields, text areas, and dropdown menus
From your documents, fills in:
* Company name and description
* Founder details and equity split
* Product/market fit explanation
For essay questions like "What's new about your solution?", synthesizes answers from your pitch deck
Reviews all entries, clicks submit, and confirms submission
## Using the Secure Browser Session
When Maia needs to access a platform that requires authentication, it will prompt you to establish a secure browser session.
### Starting a Secure Session
When your workflow needs access to a platform (like LinkedIn, your CRM, or accounting software), Maia will notify you that authentication is needed.
Maia presents an interactive button that opens the website directly in Maia's secure browser. Simply click to navigate to the login page.
Enter your credentials in the secure browser window, just as you would normally. Your login details are air-gapped from the AI system.
Once logged in, the session is saved. You can close the browser window, and Maia can begin using this authenticated session to execute your workflow.
When the workflow runs, Maia's browser picks up the session and continues from where you left off, already logged in and ready to work.
You only need to log in once per platform. The session remains active until you manually log out from that platform yourself.
## Security & Privacy: Your Credentials Stay Safe
Granting an AI access to your accounts requires trust. We built Maia's browser environment with a primary goal: absolute security for your login details.
### The Zero-Trust Approach
In simple terms, "Zero Trust" means Maia is designed to treat everything, and everyone, as potentially insecure. This is the opposite of the old way, where once you were logged in, everything was trusted.
**With Maia:**
* Maia is trusted to do the work.
* Maia is never trusted with your secrets (passwords).
This core principle means your sensitive information is isolated from the AI processes.
### How Your Login Details Are Isolated (The Air-Gap)
Think of your password and the AI as being separated by a solid wall, an "air gap." They can never touch.
1. **You Log In (Your Action)**: When a task requires authentication (like accessing your LinkedIn or your accounting portal), Maia will ask you to log in via a secure, private browser window. You type the credentials into the web page, just like you normally would.
2. **Your Credentials Never Reach Our Servers**: Your username and password are entered directly into the website you're logging into. They never pass through Maia's systems and are never stored anywhere on our servers.
3. **The Session Token (The Access Pass)**: Once you successfully log in, the website issues a session token and cookies, think of these as digitally "signed documents" that grant temporary access rights. These are the only things we store.
4. **Maia Works with the Access Pass**: When Maia runs a workflow, it uses these session tokens and cookies to access the website on your behalf. The AI agents enter the website already authenticated, ready to work, just like when you return to a website and find yourself still logged in.
5. **You Control the Access**: The session remains active until you manually log out from that platform yourself. When you log out, those "access passes" are revoked, and Maia can no longer access that account.
**Crucial Guarantee**: The AI models that execute the workflow only see the screen content (the visual data, the dashboard, the reports). It is technically impossible for the AI to access or read your actual login credentials, as those credentials were never passed to the system in the first place.
### Ethical Use
Maia's power comes with responsibility:
* **Don't spam**: Automate outreach, but keep it reasonable
* **Don't abuse**: Respect rate limits and fair use
* **Don't violate privacy**: Only access data you're authorized to see
* **Don't harm**: Avoid actions that could overload or harm websites
Learn how Maia creates professional documents from research
# Live Web Intelligence
Source: https://docs.modularmind.app/capabilities/web-intelligence
How Maia searches and gathers up-to-date information from the internet
## Live Web Intelligence (Search)
Most AI models are limited by what they learned during their training, which might be months or even years old. Maia breaks these limits by connecting directly to the live internet.
When you ask Maia a question that requires current knowledge, like "What is the latest news on AI regulations?" or "Find the current CEO of this company", it doesn't rely on memory. It performs a real-time investigation.
### Simultaneous Multi-Phrase Search
A human researcher usually searches for one thing at a time: they type a query, scan the results, refine the query, and try again. This is slow and linear.
Maia is a parallel thinker. It can generate and execute multiple search phrases simultaneously to gather a complete picture in seconds.
**Example:** If you ask, "How does our product's pricing compare to our top 3 competitors?"
* **Human approach:** Search Competitor A, write down price. Search Competitor B, write down price. Search Competitor C...
* **Maia's approach:** Maia simultaneously launches searches for "Competitor A pricing", "Competitor B pricing", and "Competitor C pricing".
It retrieves results from across the web instantly, filters for the most relevant snippets, and synthesizes the up-to-date information into a single clear answer. This ensures your automations are always powered by the freshest data available.
### Location & Maps Search
Maia's intelligence also extends to the physical world. It can perform Maps searches to retrieve location-specific information.
You don't need to do anything special to activate this, just ask a question in plain language. Whether you need to find nearby businesses, analyze region-based data, or understand the context of a specific place, Maia handles it automatically.
**How it works:** If you ask Maia to "Find the top-rated Italian restaurants within walking distance of the convention center," it will research the area, analyze the relevant details (like ratings, distance, and hours), and return accurate, actionable insights just like it does with standard web search.
Learn how Maia can actually use websites like a human
# Website & App Generation
Source: https://docs.modularmind.app/capabilities/website-app-generation
Building single-page applications, dashboards, and landing pages instantly
## From Idea to Live Website
Maia can build functional web applications and deploy them instantly — no server setup, no hosting headaches.
Tell Maia what you need, and within minutes you'll have a **live, publicly accessible URL** to share.
## What Maia Can Build
Product launches, marketing campaigns
Data visualization, analytics, KPI monitoring
Admin panels, forms, calculators, converters
Interactive web applications with complex functionality
Personal sites, project showcases, team pages
Quick mockups for concept validation
## Technology Stack
Maia builds modern, responsive web applications using:
### Frontend Technologies
* **HTML5** - Semantic structure
* **CSS3** - Modern styling and animations
* **JavaScript** - Interactive functionality
* **Responsive Design** - Mobile, tablet, desktop optimization
## Instant Hosting
Every Maia-built website is:
✅ **Automatically hosted** - No server setup required\
✅ **Publicly accessible** - Get a URL immediately\
✅ **HTTPS secured** - SSL certificates included\
✅ **Fast CDN delivery** - Global edge network\
✅ **No maintenance** - Updates and scaling handled automatically
## Building Process
### Describe Your Vision
```
Create a landing page for my bakery with:
- Hero section with our photo
- Menu items with prices
- Contact form
- Google Maps embed of our location
Use warm, inviting colors.
```
```
Build a sales dashboard that:
Layout:
- Top row: 4 KPI cards (Revenue, Deals Closed, Pipeline, Conversion Rate)
- Middle: Line chart showing monthly revenue trend
- Bottom: Table of recent deals with columns for
Client, Value, Stage, Owner
Data Source: Pull from our CRM API (I'll provide endpoint)
Styling: Corporate blue (#1E3A8A) with dark mode option
Features:
- Refresh button to reload data
- Date range filter
- Export to CSV button
```
```
Build something like airbnb.com's homepage but for
co-working spaces. Users should be able to:
- See featured spaces with photos
- Filter by city and amenities
- Click to see details
Don't need booking functionality yet, just the browsing
experience.
```
### Design Customization
Maia offers three levels of design control:
**How it works:**
```
"Create a modern dashboard with a professional look"
```
**Result:** Maia chooses colors, fonts, layout based on best practices
**Best for:** Quick prototypes, internal tools, getting started fast
**How it works:**
```
Colors: Blue (#1E3A8A) and white
Font: Sans-serif, clean and modern
Style: Minimalist with lots of whitespace
```
**Result:** Maia designs within your constraints
**Best for:** Branded content, specific aesthetic preferences
**How it works:**
```
"Use this exact CSS for the header:
.header {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
padding: 2rem;
box-shadow: 0 4px 6px rgba(0,0,0,0.1);
}
And implement this card component: [paste HTML code]"
```
**Result:** Maia incorporates your exact code
**Best for:** Developers, precise visual requirements, reusing existing components
## Integration with Maia Workflows
### Apps That Trigger Workflows
Build interfaces that launch your Maia automations:
**Example:**
```
Create a dashboard with a button labeled "Generate Monthly Report"
When clicked, trigger my "Monthly Report" workflow and display
the PDF when it's ready.
```
**Result:** A simple UI that non-technical team members can use to run complex automations.
## Personalized Web Pages at Scale
Maia excels at generating highly personalized, stylized web pages for your customers, leads, or stakeholders. Whether you need custom proposals, personalized reports, or tailored presentations, Maia can:
* **Accept instructions and data** - Provide templates, brand guidelines, and recipient information
* **Gather additional context** - Automatically research and incorporate relevant details
* **Generate at scale** - Create hundreds of unique, personalized pages efficiently
* **Deliver professional results** - Produce polished, shareable pages ready for distribution
Each page can be customized with recipient-specific content, data visualizations, and styling that matches your brand identity — making every interaction feel personal and professional.
See comprehensive examples of building websites, apps, and personalized pages with Maia:
## Limitations
**What Maia builds:**
* Frontend web applications
* Single-page applications (SPAs)
* Static sites with API integrations
**What Maia doesn't build (yet):**
* Full backend services (databases, authentication servers)
* Mobile native apps (iOS, Android)
* Complex multi-user real-time apps (like Slack clones)
Discover how Maia can generate and run code for data analysis, calculations, and visualizations
# FAQ
Source: https://docs.modularmind.app/faq
Understanding plan tiers, workflow limits, and access
Yes! Maia is available on both iOS and Android. The mobile app gives you the full Maia experience — kick off and monitor workflows on the go, chat with your executives in real time, get notified when work completes, and access your entire workspace from anywhere.
Yes, upgrade anytime (takes effect immediately). Downgrade at end of billing period.
You can't deploy new workflows but existing ones keep running. Upgrade to add more.
Workflow limits are concurrent (not cumulative), so there's nothing to "roll over."
Not currently, but Enterprise plans offer unlimited workflows. Contact sales for custom solutions.
Not yet, but planned for future release. Currently, choose a plan that fits your needs.
***
**Ready to get started?**
Try any plan free for 14 days
Compare plans and pricing
Discuss Enterprise needs
# Custom Actions
Source: https://docs.modularmind.app/integrations/custom-actions
Teaching Maia new tricks with custom HTTP requests and API integrations
## Extend Maia's Capabilities
While Maia has built-in actions, as well as 3rd party connection actions, you can teach it new ones by defining custom API calls. Once created, Maia remembers and uses them intelligently in future workflows.
Custom actions let you integrate any web service with a RESTful API, from proprietary internal tools to niche third-party services.
## When to Create Custom Actions
Your company's custom CRM, inventory system, or proprietary database
Industry-specific tools not in Maia's built-in library
Services behind authentication
Older systems with custom APIs that need specific formatting
## How It Works
Custom actions are organized into **Platforms** and **Actions** you can define. Each platform represents a service or system, and each platform can have multiple actions.
The key to making custom actions work effectively is using **natural language descriptions**. Maia reads these descriptions to understand what each platform and action does, when to use them, and how to fill in the required parameters automatically during workflow execution.
## Managing Platforms
The Actions menu displays all your defined platforms. You can:
* **Add Platform**: Create a new platform to group related actions
* **Enable/Disable Platforms**: Control whether Maia can discover and use actions from a specific platform
* **Edit Platform**: Update the platform name and description
## Creating and Managing Actions
Within each platform, you can define multiple actions. Each action represents a specific HTTP request (API call) that Maia can execute.
### Action Controls
For each action, you can:
* **Enable/Disable**: Toggle whether Maia can use this action in workflows
* **Edit Action**: Modify the action configuration
* **Delete Action**: Remove actions you no longer need
When an action is enabled, Maia can automatically search for and use it when appropriate based on the action's description and the workflow context.
## Adding an HTTP Request Action
To create a new custom action, click **Add Action** within a platform. Then choose **Add an HTTP Request Action**.
When you add a new HTTP request action, you configure the following:
### Basic Information
**Title**: The name of your action (e.g., "Read Database Entry", "Get User Info")
**Description**: A brief explanation of what the action does. This is crucial — Maia uses this description to determine when this action is relevant to use in a workflow.
### Request Configuration
**Method**: Choose the HTTP method for your request:
* `GET` - Retrieve data
* `POST` - Create or send data
* `PATCH` - Update existing data
**URL**: The API endpoint (e.g., `https://api.example.com/data`)
**Use Dynamic Path**: Toggle this if parts of your URL need to be dynamically filled based on parameters
### Parameters
You can define four types of parameters:
#### Query Parameters
Parameters sent in the URL query string (e.g., `?user_id=123&status=active`)
#### Headers
HTTP headers sent with the request (e.g., authentication tokens, content types)
For each header, you can specify:
* **Name**: The header name (e.g., `x-maia-secret`)
* **Required**: Whether this header must be included
* **Constant Value**: Set a fixed value so Maia doesn't need to determine it dynamically
#### Path Parameters
Dynamic segments in the URL path (enabled when "Use Dynamic Path" is toggled)
#### Body Parameters
Data sent in the request body (for POST and PATCH requests)
### Parameter Configuration
For each parameter (query, header, path, or body), you can define:
**Name**: The parameter name
**Description**: Optional natural language explanation of the parameter. This helps Maia understand when and how to use it, what values are appropriate, and where to get the values from in the workflow context.
**Required**: Toggle whether this parameter is mandatory for the request
**Constant Value**: If a parameter needs a fixed value, enter it here. This is useful for parameters that are required but don't change (like API keys, version numbers, or fixed configurations). When a constant value is set, Maia uses this value automatically without needing to determine it.
Constant values are securely encrypted and remain completely private from the AI models powering Maia (Gemini, GPT, Claude) and their providers (Google, OpenAI, Anthropic). This ensures safe storage of sensitive information like API keys.
## Testing Actions
Before saving an action, you can test it to ensure it works as expected:
1. Click **Test Request** to open the test interface
2. The test screen shows your configured request details:
* Method (GET, POST, PATCH)
* URL
* Headers
* Body Parameters
3. Manually provide example values for the parameters
4. Click "Send Request" to execute the test
5. View the response on the right side to verify the API call works correctly
This testing capability helps you validate your action configuration before Maia attempts to use it in an actual workflow.
## Best Practices
Write detailed, natural language descriptions for platforms, actions, and parameters. The better the description, the more accurately Maia will use them.
Set constant values for parameters that never change (API keys, fixed IDs, version numbers) to simplify Maia's job.
Organize actions into logical platforms to keep your integrations organized and easier to manage.
Always test your actions with example data to catch configuration errors early.
Learn how to deploy your workflows to run automatically
# Data Sources
Source: https://docs.modularmind.app/integrations/data-sources
Access industry-leading data intelligence directly in your workflows
## Native Data Sources
Maia natively supports a suite of powerful Data Sources, enabled with a single click. Instead of manually searching and scraping, you can instruct Maia to pull precise information from the most trustworthy providers for lead generation, SEO, and market research.
This lets you focus on strategy and execution while Maia handles the heavy lifting of data collection with programmatic accuracy.
## How Data Sources Work
### Just Ask
You don't need to configure anything or select which data source to use. Just describe what you need in natural language, and Maia automatically pulls from the right source.
```
"Find marketing managers at SaaS companies in California
with 50-200 employees and get their contact information"
```
1. Identifies this as a lead generation request
2. Searches for matching profiles by title, company size, and location
3. Enriches results with verified emails and phone numbers
4. Returns structured lead data ready for outreach
### Saved to Your Workspace
Unlike Native Connections which access data transiently during workflow execution, Data Sources save retrieved data to your workspace. You can view, export, and reuse this data across multiple workflows at any time.
## Example Workflows
**What you ask:**
```
"Find CTOs at fintech startups in New York that raised
Series A funding this year. Get their email addresses."
```
**What happens:**
Maia searches for people matching your criteria—filtering by job title,
industry, location, and funding stage—then enriches the results with
verified contact information ready for your sales outreach.
**What you ask:**
```
"Analyze the keyword 'AI automation tools' and find related
keywords with high search volume but low competition."
```
**What happens:**
Maia retrieves comprehensive keyword metrics including search volume,
keyword difficulty, CPC, and search intent. It then discovers related
keyword opportunities and filters for high-potential targets.
**What you ask:**
```
"Find the top YouTube videos about product-led growth
from the last month and summarize what topics they cover."
```
**What happens:**
Maia searches YouTube for relevant videos, extracts their metadata
and full transcripts, then analyzes the content to identify common
themes and talking points.
**What you ask:**
```
"Compare search interest for 'remote work' vs 'hybrid work'
over the past year and identify rising related topics."
```
**What happens:**
Maia queries trend data to show interest over time, compares the
terms across regions, and surfaces related queries that are
gaining momentum.
Learn how Maia dynamically selects the right tools for each task
# Intelligent Action Use
Source: https://docs.modularmind.app/integrations/intelligent-actions
How Maia dynamically discovers and uses the right tools for your workflows
## Smart Tool Selection
Maia doesn't require you to manually specify which connections to use. It intelligently selects the right actions based on your intent.
Think of it like talking to a skilled assistant who knows which tools to reach for — you describe the outcome, not the implementation.
## Suggested Actions vs. Available Actions
### Suggested Actions
When building a workflow, Maia **suggests** actions for AI agents based on your request:
**Your Request:**
```
"Retrieve Instagram posts from my feed, generate a similar
image for posting,and email me it as an attachment."
```
**Maia's suggested actions for the prompt:**
* Generate Image: Create an image using Nano Banana Pro
* Gmail: Send email with attachment
These suggestions appear in the workflow visualization before execution.
### Available Actions (Dynamic Discovery)
During execution, if Maia encounters a situation not covered by the plan, it **searches its action library** for solutions:
**Scenario:**
* Planned action: "Read the report from Dashboard"
* Problem: Dashboard requires 2FA code
* **Dynamic discovery**: Maia searches for "Get 2FA code" actions
* **Finds**: Gmail action to read 2FA code from email
* **Executes**: Reads code, completes authentication, proceeds with download
**Result:** Workflow adapts without manual intervention
Learn how to teach Maia new actions with custom API calls
# Native Connections
Source: https://docs.modularmind.app/integrations/native-connections
Connecting Maia to 3rd party essential business tools
## Seamless Tool Connection
Maia integrates natively with popular business applications, allowing workflows to read, write, and manage data across your existing tools.
You don't need to manually configure "nodes" or "connectors." Just tell Maia what you need, and it automatically uses the right connection.
## How Connections Work
### One-Time Authorization
When you mention a tool for the first time:
```
"Check my Gmail for messages from customers"
```
Maia prompts you to authorize access via standard OAuth
You approve the specific permissions Maia needs (e.g., "Read Gmail messages")
Your authorization is encrypted and stored securely
Maia automatically uses the saved authorization — no repeated logins needed
Optionally, you can have a look at available connections through Connections settings in your Maia workspace, and connect your accounts there as well.
### Automatic Action Selection
You don't specify action use, Maia infers what you need:
```
"Read emails from support@customer.com,
add a row to the 'Customer Feedback' Google Sheet
with the sender, subject, and date."
```
1. Gmail: Search inbox
2. Gmail: Filter by sender
3. Gmail: Extract subject and date
4. Google Sheets: Find 'Customer Feedback' sheet
5. Google Sheets: Append new row
## Privacy & Data Access
**Important:** Maia only accesses data needed for your specific workflows. The AI model processes data transiently — nothing is stored permanently without your explicit instruction.
### What Maia Can See
* Only data from authorized accounts
* Only during workflow execution
* Only for the specific task requested
### What Maia Cannot See
* Data from non-authorized accounts
* Information outside the workflow scope
* Historical data after workflow completes
### Your Controls
* Revoke access anytime
* Audit responses show what was accessed
Learn how Maia dynamically selects the right tools for each task
# AI Executives
Source: https://docs.modularmind.app/intro/ai-executives
Understanding the AI executive paradigm that powers Maia
## Beyond Agents
The dominant approach to AI agents today has a structural ceiling: a single context window accumulates every instruction, tool call, and intermediate output until it collapses under its own weight. Agents drift. Priorities blur. Long-horizon work fails because the architecture forces it to hold too much at once with no principled way to offload it.
When the industry talks about agents, it means single-purpose tools — a presentation generator, a support triage bot, a scheduling assistant. That framing puts the burden back on the human to know which tool to call, when, and how to stitch outputs together. You're still the one running the operation.
An AI executive operates at a higher level entirely.
## What Makes an Executive
Three things distinguish an AI executive from an AI agent:
### Workflows
A workflow is a modular SOP — a process designed in detail, deployed reliably, and refined independently. Unlike static documents, workflows compose hierarchically: they can be called by an executive, return results upward, and be modified without touching anything else in the system. This is what makes complex, repeatable operations possible at the executive level.
See how workflows are deployed and triggered
### Hierarchical Memory
An executive's effectiveness compounds over time. Memory acts as a context fabric across decoupled workflows — the executive knows what happened last week, what's been decided, what constraints apply. Not because it was told again, but because it retains it. Without this, workflows operate in isolation, repeating discovery instead of building on prior results.
See how memory works in Maia
### Multi-Agent Coordination
An executive doesn't execute alone. It coordinates — running specialized agents in parallel, spawning multiple instances where throughput demands it, and synthesizing results from across the swarm. The executive is the coordination layer, not the execution primitive.
## The Executive Paradigm
The human defines the goal and the standards. The executive handles the complexity in between. Operations that used to require entire teams — coordinated, context-aware, sustained across time — are now within reach of a single executive.
For the full architectural deep-dive on how AI executives work and why they surpass traditional agent frameworks, read [Introducing AI Executives](https://www.modularmind.app/blog/introducing-ai-executives/) on the ModularMind blog.
Learn about the core execution loop that powers autonomous work
# How Maia Works
Source: https://docs.modularmind.app/intro/how-maia-works
Understanding the core execution loop that powers autonomous work
## The Core Loop
Maia follows a simple but powerful four-step process to turn your goals into completed work:
Tell Maia what you want to accomplish in plain language. Be as detailed as you'd be when explaining a task to a colleague.
```
"Research the top 5 CRM tools for a bakery business,
compare their pricing and features, and create a
presentation I can share with my team."
```
Maia analyzes your request and creates a comprehensive plan. It identifies:
* What information needs to be gathered
* Which tools and integrations to use
* How to break the work into parallel tasks
* The optimal sequence of operations
Your strategy becomes a visual workflow: a step-by-step diagram showing exactly what Maia will do. You can:
* See the plan before it runs
* Understand the logical flow
* Make adjustments if needed
* Learn how the process works
Your executive coordinates multiple specialized agents working in parallel to complete the workflow. You can:
* Watch the work happen in real-time
* See results as they're generated
* Deploy for future automation
## From Manual to Automated
The key to getting great results from Maia is describing your current manual process:
```text Good Example theme={null}
"I usually visit G2.com and search for CRM software.
Then I open each product page, scroll down to find
pricing, and copy the details into a spreadsheet.
After comparing 5-10 options, I create a presentation
with screenshots and my notes."
```
```text Also Works theme={null}
"Compare popular CRM tools for bakeries and create
a presentation with pricing and features."
```
The more detail you provide about your current workflow, the better Maia can optimize and automate it.
## Intelligent Adaptation
Maia doesn't just follow a script—it adapts during execution:
* **Dynamic Decision Making**: Adjusts approach based on what it discovers
* **Error Recovery**: Handles unexpected situations automatically
* **Parallel Processing**: Scales up agents when beneficial
See Maia in action with a simple example
# Quick Start Guide
Source: https://docs.modularmind.app/intro/quick-start
Get started with Maia in minutes with this hands-on example
## Your First Workflow
Let's walk through a practical example that showcases Maia's capabilities: researching CRM tools and creating a comparison document.
Open Maia and describe what you need:
```
Research the top 5 CRM tools for a bakery business
and create a comparison PDF with pricing and key features.
```
Maia will acknowledge your request and start building a strategy.
Maia presents a visual workflow showing:
* **Web Research**: Search for "best CRM for bakeries"
* **Data Collection**: Visit product pages and gather information
* **Comparison Analysis**: Structure findings into a table
* **Document Generation**: Create a styled PDF presentation
You'll see each step connected in a logical flow.
If you have already provided enough detail, Maia will automatically run the workflow. Otherwise, you can confirm by saying something like:
```
Looks great! Run it.
```
or using the "Run Workflow" button, located bottom left corner of the board.
Maia will then begin execution. You'll see real-time updates as each agent:
* Searches multiple sources simultaneously
* Extracts relevant information
* Compiles comparison data
* Generates your PDF
Within minutes, you'll receive:
* A polished PDF comparison document you can download
* Source links for verification
* A workflow you can run again, or deploy with routine, webhook, or action triggers
## What Just Happened?
Behind the scenes, Maia:
1. **Analyzed** your request to understand the goal
2. **Planned** a multi-step research and creation workflow
3. **Instructed** multiple agents to work in parallel
4. **Searched** the web with multiple query strategies
5. **Extracted** structured data from various sources
6. **Generated** a professional document with proper formatting
All without you needing to:
* Write any code
* Configure integrations
* Manage multiple tools
* Copy and paste between applications
## Try These Next
```
Every Monday, check my Gmail for unread messages
from customers, summarize them, and add to a
Google Sheet with priority levels.
```
```
Monitor competitor websites weekly and create
a report on any pricing changes or new features.
```
```
Analyze my top 3 performing Instagram posts from last month,
generate 5 new bakery content ideas based on those themes,
create captions with trending hashtags.
```
```
Extract invoice data from PDFs in my Drive folder
and populate a Google Sheet with organized records.
```
## Tips for Success
Instead of "research CRM tools," say "research CRM tools for a small bakery with 10 employees that needs inventory tracking."
Tell Maia how you currently do the task manually. This helps it understand the nuances and optimize the workflow.
Begin with a straightforward workflow. Once you see how Maia handles it, you can add complexity.
Run a workflow once manually to verify results. Then deploy it for scheduled automation.
**On the go?** Maia is available on iOS and Android. Kick off workflows, chat with your executives, and get notified when work completes — all from your phone.
Dive deeper into how to interact with Maia effectively
# What is Maia?
Source: https://docs.modularmind.app/intro/what-is-maia
Meet your AI executives — operational teammates that plan, coordinate, and execute complex work
## More Than an AI Agent
Maia is an **AI executive platform** — build and customize AI executives that understand operations end-to-end, collaborate with teams, and run complex workflows across your company. Unlike chatbots that answer questions or single-purpose agents that handle one task, Maia executives plan, coordinate, and execute sustained operational work.
## The Multi-Model Advantage
Every AI executive runs on a massively multi-agent system that orchestrates multiple state-of-the-art AI models to deliver the best results.
Under the hood, Maia coordinates agents powered by **GPT 5.6 Sol**, **Claude Fable 5**, and **Gemini 3.1 Pro**, assigning each task to the model best suited for the job.
Each task is automatically routed to the agent that will perform it best, giving your executives the combined strengths of the world's leading AI models.
## Massively Multi-Agent Coordination
At its core, Maia uses **massively multi-agent chain-of-thought** — your executives don't execute alone, they coordinate:
1. **Thinks**: Analyzes your request and understands the goal
2. **Plans**: Breaks down complex tasks into semantic steps
3. **Delegates**: Assigns each sub-task to specialized agents
4. **Executes**: Coordinates multiple agents working in parallel
This means your executives can handle operations that would normally require a team of specialists — from research and data analysis to document creation and web automation.
## Built for Everyone
Whether you're a small business owner automating repetitive tasks or a developer building custom integrations, Maia adapts to your level of technical expertise:
* **No-code interface** for business users
* **Visual workflows** for process optimization
* **Custom AI executives** for different operational domains
* **Flexible deployment** with webhooks, schedules, and triggers
Learn what makes AI executives different from traditional AI agents
# The Conversation Interface
Source: https://docs.modularmind.app/planning/conversation-interface
How to communicate effectively with your AI teammate
## Talking to Maia
Maia's interface is designed for natural conversation. You don't need to learn special commands or syntax — just describe what you need in your own words.
## How the Interface Works
Everything happens in the same chat interface. When you describe what you need, Maia collaborates with you through conversation to understand your requirements and define the workflow. As Maia starts planning your solution, a board view appears on the right side, visualizing the workflow steps.
**During the conversation, you can:**
* Discuss requirements and goals
* Develop and refine strategy
* Review the visual workflow on the board
* Ask questions and provide clarifications
* Request adjustments based on your needs
* Approve the plan when ready
**Example conversation:**
```
You: I need to track customer feedback from emails
Maia: I can help with that. Should I:
1. Monitor a specific email inbox?
2. Look for particular keywords or senders?
3. Store the feedback in a spreadsheet?
You: Yes, monitor support@mybakery.com for emails
with "feedback" or "suggestion" in the subject
Maia: Got it. I'll create a workflow that checks
that inbox, filters by those keywords, and
logs entries to a Google Sheet. Would you like
to run this once or schedule it regularly?
```
**Once you approve the plan:**
* The workflow executes in real-time
* You see progress updates in the chat
* The board view shows which agents are working on what
* Results are generated and delivered
* Errors are handled and communicated
**During execution, you can:**
* Monitor progress in real-time
* Pause execution if needed
* Provide additional context mid-run
## Describing Your Manual Process
The best way to get Maia to automate something is to describe how you currently do it manually.
💬 *Right now, I go to our G2.com company page every Friday.
I scroll through new reviews, copy any that mention
specific features, paste them into a Google Doc, and
then email the doc to our product team with a summary
of common themes.*
✅ **Maia understands:**
* Where to find information (G2.com)
* What to look for (feature mentions)
* Where to store it (Google Doc)
* Who needs it (product team via email)
* When to do it (weekly on Friday)
💬 *I need to monitor our G2 reviews and share relevant
feedback with the product team weekly.*
✅ **Maia will ask:**
* Which reviews are "relevant"?
* How should the feedback be formatted?
* Any specific team members to notify?
* Preferred delivery method?
## Effective Communication Tips
If Maia seems confused or the results aren't what you expected:
Show Maia what the ideal output looks like:
"Here's an example of the format I want:
.."
Split complex requests into smaller steps:
"First, just show me the research results.
Then we'll create the document."
Explain the business context:
"I need this for a board presentation, so it
should be formal and data-focused."
Learn how to edit individual step instructions
# Prompts
Source: https://docs.modularmind.app/planning/prompts
Understanding the building blocks of workflow steps
## What is a Prompt?
Every step in your workflow is powered by a **prompt** — detailed instructions that tell an AI agent exactly what to do.
Think of a prompt as a **job description** for an AI agent. It contains all the context, instructions, and expectations for completing one specific task.
## The Problem with Traditional Prompts
Traditionally, prompts are monolithic blocks of text:
*You are a research assistant. Search the web for CRM tools
suitable for small bakeries with 10-15 employees. Focus on
tools with inventory management, customer communication, and
order tracking. Budget is under \$500/month. Extract pricing,
key features, user ratings, and pros/cons. Format the results
as a structured table. Prioritize tools with high ratings and
good integration capabilities...
\[continues for several paragraphs]*
**Challenges:**
* Hard to read and understand
* Difficult to modify one part without breaking others
* Can't reuse pieces across workflows
* Overwhelming to edit for non-technical users
## Maia's Modular Approach
Maia breaks prompts into sections — discrete, editable sections that serve specific purposes. Each section can be expanded to view and edit its full content.
**Example:** Research prompt broken into sections
You are an expert B2B SaaS researcher with deep expertise in small business CRM tools, specifically focused on understanding the unique requirements of the bakery industry. Your background includes 10+ years analyzing software solutions for food service businesses, with particular emphasis on inventory management and customer relationship systems. You understand the challenges small bakeries face, including managing perishable inventory, handling custom orders, tracking customer preferences for special occasions, maintaining communication with both retail and wholesale clients, and balancing operational efficiency with personalized service. Approach this research with the mindset of a bakery owner who needs practical, cost-effective solutions that can scale with business growth.
Your objective is to find and compare the top 5 CRM tools available in the market that are specifically well-suited for small bakery operations. Focus your research on reputable sources including G2, Capterra, TrustRadius, Software Advice, and verified user reviews to ensure comprehensive coverage. Prioritize tools that have demonstrated success in the food service or retail bakery sector, with verified case studies or testimonials from similar businesses. Evaluate each tool's ease of implementation, learning curve for staff, integration capabilities with common bakery point-of-sale systems, and mobile accessibility for on-the-go management. Additionally, assess the quality and responsiveness of customer support, as this is critical for small businesses without dedicated IT staff.
The budget must stay under \$500 per month for a team size of 10-15 employees, with preference given to solutions offering flexible user-based pricing models. The CRM solution must include inventory management capabilities that can handle perishable goods with expiration date tracking, robust customer communication features supporting email, SMS, and automated birthday or special occasion reminders, and comprehensive order tracking functionality for both one-time and recurring orders. The system should support custom product configurations for specialty cakes and pastries, integrate with at least two major payment processors, and provide mobile apps for both iOS and Android platforms. Data security and GDPR compliance are mandatory requirements, and the solution should offer data export capabilities to prevent vendor lock-in.
Present the results as a structured table sorted by rating (highest first), with each tool occupying one row. Each entry should include the tool name with an official website link, detailed pricing information breaking down costs by user tier and any additional fees, key features listed as bullet points with special emphasis on bakery-specific functionality, aggregate user ratings from multiple platforms with the number of reviews cited, and a balanced list of pros and cons with at least 3-4 items in each category. Include a separate column for "Best For" that describes the ideal bakery profile for each tool (e.g., "Best for wholesale-focused bakeries" or "Best for retail bakeries with multiple locations"). Add a final summary row that provides an overall recommendation based on the most common bakery business model.
## Benefits of Modular Prompts
### **Surgical Editing**
Change one aspect without rewriting everything. Here's the same CRM research prompt with only the budget requirement modified:
You are an expert B2B SaaS researcher with deep expertise in small business CRM tools, specifically focused on understanding the unique requirements of the bakery industry. Your background includes 10+ years analyzing software solutions for food service businesses, with particular emphasis on inventory management and customer relationship systems. You understand the challenges small bakeries face, including managing perishable inventory, handling custom orders, tracking customer preferences for special occasions, maintaining communication with both retail and wholesale clients, and balancing operational efficiency with personalized service. Approach this research with the mindset of a bakery owner who needs practical, cost-effective solutions that can scale with business growth.
Your objective is to find and compare the top 5 CRM tools available in the market that are specifically well-suited for small bakery operations. Focus your research on reputable sources including G2, Capterra, TrustRadius, Software Advice, and verified user reviews to ensure comprehensive coverage. Prioritize tools that have demonstrated success in the food service or retail bakery sector, with verified case studies or testimonials from similar businesses. Evaluate each tool's ease of implementation, learning curve for staff, integration capabilities with common bakery point-of-sale systems, and mobile accessibility for on-the-go management. Additionally, assess the quality and responsiveness of customer support, as this is critical for small businesses without dedicated IT staff.
The budget must stay under **\$500** per month for a team size of 10-15 employees, with preference given to solutions offering flexible user-based pricing models. The CRM solution must include inventory management capabilities that can handle perishable goods with expiration date tracking, robust customer communication features supporting email, SMS, and automated birthday or special occasion reminders, and comprehensive order tracking functionality for both one-time and recurring orders. The system should support custom product configurations for specialty cakes and pastries, integrate with at least two major payment processors, and provide mobile apps for both iOS and Android platforms. Data security and GDPR compliance are mandatory requirements, and the solution should offer data export capabilities to prevent vendor lock-in.
Present the results as a structured table sorted by rating (highest first), with each tool occupying one row. Each entry should include the tool name with an official website link, detailed pricing information breaking down costs by user tier and any additional fees, key features listed as bullet points with special emphasis on bakery-specific functionality, aggregate user ratings from multiple platforms with the number of reviews cited, and a balanced list of pros and cons with at least 3-4 items in each category. Include a separate column for "Best For" that describes the ideal bakery profile for each tool (e.g., "Best for wholesale-focused bakeries" or "Best for retail bakeries with multiple locations"). Add a final summary row that provides an overall recommendation based on the most common bakery business model.
You are an expert B2B SaaS researcher with deep expertise in small business CRM tools, specifically focused on understanding the unique requirements of the bakery industry. Your background includes 10+ years analyzing software solutions for food service businesses, with particular emphasis on inventory management and customer relationship systems. You understand the challenges small bakeries face, including managing perishable inventory, handling custom orders, tracking customer preferences for special occasions, maintaining communication with both retail and wholesale clients, and balancing operational efficiency with personalized service. Approach this research with the mindset of a bakery owner who needs practical, cost-effective solutions that can scale with business growth.
Your objective is to find and compare the top 5 CRM tools available in the market that are specifically well-suited for small bakery operations. Focus your research on reputable sources including G2, Capterra, TrustRadius, Software Advice, and verified user reviews to ensure comprehensive coverage. Prioritize tools that have demonstrated success in the food service or retail bakery sector, with verified case studies or testimonials from similar businesses. Evaluate each tool's ease of implementation, learning curve for staff, integration capabilities with common bakery point-of-sale systems, and mobile accessibility for on-the-go management. Additionally, assess the quality and responsiveness of customer support, as this is critical for small businesses without dedicated IT staff.
The budget must stay under **\$200** per month for a team size of 10-15 employees, with preference given to solutions offering flexible user-based pricing models. The CRM solution must include inventory management capabilities that can handle perishable goods with expiration date tracking, robust customer communication features supporting email, SMS, and automated birthday or special occasion reminders, and comprehensive order tracking functionality for both one-time and recurring orders. The system should support custom product configurations for specialty cakes and pastries, integrate with at least two major payment processors, and provide mobile apps for both iOS and Android platforms. Data security and GDPR compliance are mandatory requirements, and the solution should offer data export capabilities to prevent vendor lock-in.
Present the results as a structured table sorted by rating (highest first), with each tool occupying one row. Each entry should include the tool name with an official website link, detailed pricing information breaking down costs by user tier and any additional fees, key features listed as bullet points with special emphasis on bakery-specific functionality, aggregate user ratings from multiple platforms with the number of reviews cited, and a balanced list of pros and cons with at least 3-4 items in each category. Include a separate column for "Best For" that describes the ideal bakery profile for each tool (e.g., "Best for wholesale-focused bakeries" or "Best for retail bakeries with multiple locations"). Add a final summary row that provides an overall recommendation based on the most common bakery business model.
Notice how only the **Requirements & Constraints** section is changed (budget from \$500 to \$200), and the rest of the prompt stays intact.
### **Clarity and Organization**
Instead of hunting through paragraphs, you can expand and edit relevant sections:
* Need to change what data is collected? → Edit **Task Definition**
* Want different formatting? → Edit **Output Format**
* Add requirements? → Edit **Requirements & Constraints**
Learn what Maia can do inside those workflow steps
# Security & Privacy
Source: https://docs.modularmind.app/security-privacy
How Maia protects your data, credentials, and business information
## Security First
Maia handles sensitive business data and credentials. Security and privacy are fundamental to every aspect of the platform.
Your data is **your data**. Maia processes it only as needed for your workflows and **never** uses it to train AI models or share with third parties.
## Data Handling Principles
### 1. Isolation
**Your data is completely isolated:**
* Each account operates in a separate, encrypted environment
* No data leakage between customers
* Even ModularMind staff cannot access your data without explicit permission
### 2. Transient Processing
**Data is not stored permanently:**
* AI models process data **in memory** during workflow execution
* Only results are stored in your workspace
* Unless you explicitly ask Maia to send it to an external database, processed data is discarded after execution
### 3. No Training Data
**Critical:** Your data is **NEVER** used to train AI models.
When Maia uses GPTs, Claude, or Gemini:
* Data is sent to these services for processing
* Maia uses enterprise agreements with these providers that **prohibit training** on customer data
* Your workflows and data remain confidential
## Credential Security
### Browser Session Storage
Maia's "Secure Browser" capability uses **air-gapped credentials**:
You authenticate to a website (e.g., Gmail, LinkedIn) in a secure browser session
The **session cookies** (not your password) are encrypted and stored
**Encryption:** AES-256 encryption at rest
When workflows need access, Maia uses the stored session
**AI Model Sees:** Only the rendered screen (visual pixels)
**AI Model Does NOT See:** Cookies, passwords, authentication tokens
You can revoke stored sessions anytime from your Maia settings
**Key Point:** Maia uses your authenticated session, but the AI model only sees the screen output — like watching someone use a computer, not reading their password manager.
### OAuth Tokens
When you authorize Maia to access services:
1. **Standard OAuth 2.0 flow** (same as "Login with Google")
2. **Tokens stored encrypted**
3. **Automatic refresh** when tokens expire
4. **Revocable** from service provider
## Network Security
### Infrastructure
Maia runs on enterprise-grade infrastructure:
* **AWS / GCP**: Industry-leading cloud providers
* **Private VPCs**: Isolated network environments
* **Firewalls**: Restrict access to authorized traffic only
* **DDoS Protection**: Provider-level protection
## Privacy & Compliance
### GDPR Compliance
For EU customers:
✅ **Right to Access**: Download all your data\
✅ **Right to Erasure**: Delete all data on request\
✅ **Data Portability**: Export in standard formats\
✅ **Consent Management**: Clear consent for data processing\
✅ **Data Processing Agreement**: Available for enterprise customers
## Incident Response
### Monitoring
Maia's security team monitors for:
* Unauthorized access attempts
* Unusual API activity
* Data exfiltration attempts
* Service disruptions
### Breach Notification
In the unlikely event of a security breach:
Automated systems + security team identify issue
Immediate action to stop unauthorized access
Affected customers notified within **72 hours**
Email includes:
* What happened
* What data was affected
* What we're doing about it
* What you should do
Fix vulnerability, enhance security
Public transparency report (if appropriate)
## Trust Center
For more security information:
* **[Privacy Policy](https://maia.is/privacy)**: How we handle your data
* **[Terms of Service](https://maia.is/terms)**: Legal agreement
Have questions about plans, limits, and model access? Find answers here.
# Action Triggers
Source: https://docs.modularmind.app/workflows/action-triggers
Turn deployed workflows into reusable building blocks that Maia can call within larger plans
## Workflows as Building Blocks
Action triggers let you deploy a workflow as a **reusable action** that Maia can automatically call and execute as part of larger, more sophisticated operations. Instead of rebuilding the same logic across multiple plans, you build it once and Maia uses it wherever it's needed.
With action triggers, your workflows become a **library of high-performance tools** that Maia orchestrates automatically — turning custom logic into modular building blocks.
## How Action Triggers Work
When you enable an action trigger on a deployed workflow, Maia registers it as an available action — just like actions from connections, custom platforms, and data sources. Maia can then discover and call it during planning and execution whenever it's relevant.
```mermaid theme={null}
graph LR
A[Larger Plan] -->|Calls action| B[Deployed Workflow]
B --> C[Executes Full Workflow]
C --> D[Returns Results]
D --> A
```
### Enabling an Action Trigger
Build and deploy your workflow, or open an existing deployed workflow.
Every deployed workflow comes with an action trigger. In the trigger management section, toggle it on. Maia automatically generates an action name based on your workflow's title, and the workflow becomes available as an action that Maia can call.
## What Makes This Powerful
### Composability
Maia can combine multiple workflow actions together to solve complex, multi-step problems. Each workflow action is a self-contained unit that Maia calls as needed, just like any other action in its toolkit.
### Automatic Discovery
You don't need to explicitly tell Maia which workflow actions to use. During planning, Maia discovers relevant actions based on their names and descriptions and incorporates them into its strategy automatically.
### Modular Scalability
Since each building block is consistent and editable on its own, you can improve a single workflow action and instantly upgrade every complex plan that relies on it — without modifying the plans themselves.
## Use Cases
Build a "Lead Scoring" workflow that evaluates prospects based on your exact criteria. Once deployed as an action, Maia calls it as a step within larger flows — like a full lead generation pipeline or a multi-touch warm-up sequence — scoring each lead in context without you lifting a finger.
**Example flow:**
```
Lead Generation Pipeline
├── Scrape leads from target websites
├── → Call "Lead Scoring" action for each lead
├── Filter leads with score > 70
├── Enrich high-scoring leads with LinkedIn data
└── Add qualified leads to CRM
```
Define your brand's visual identity once in a "Brand Image Generator" workflow. Maia calls it whenever visuals are needed across larger content or campaign flows — ensuring every generated image stays consistent with your brand, automatically.
**Example flow:**
```
Content Campaign Builder
├── Generate blog post from topic brief
├── → Call "Brand Image Generator" for hero image
├── → Call "Brand Image Generator" for social thumbnails
├── Format for website publication
└── Schedule social media posts
```
Package specialized data processing — like cleaning, validation, or transformation — into a workflow action. Any plan that handles raw data can call your processing workflow to apply consistent standards.
**Example flow:**
```
Monthly Reporting Pipeline
├── Pull raw data from multiple sources
├── → Call "Data Cleaner" action on each dataset
├── Merge cleaned datasets
├── Generate charts and analysis
└── Create PDF report and distribute
```
Create a compliance-checking workflow that reviews content against your policies. Maia can call it wherever content is generated — emails, reports, social posts — ensuring everything meets your standards before delivery.
**Example flow:**
```
Client Communication Flow
├── Draft response based on client inquiry
├── → Call "Compliance Review" action on draft
├── Apply recommended changes
└── Send approved response
```
## Combining with Other Triggers
Action triggers work alongside routine and webhook triggers on the same workflow. For example, a lead scoring workflow could:
* **Action trigger**: Maia calls it within a larger lead generation plan
* **Webhook trigger**: Your website form calls it directly for real-time scoring
* **Routine trigger**: It runs every morning to re-score yesterday's leads with updated data
All three triggers execute the same workflow, keeping your scoring logic consistent across every entry point.
Trigger workflows from external events and return intelligent responses
# Deployed Workflows
Source: https://docs.modularmind.app/workflows/deployed-workflows
Deploy workflows to run automatically with flexible triggers
## From Chat to Automation
Once you've perfected a workflow by watching it run, you can **deploy it to the cloud** to run automatically — no need to keep Maia open or manually trigger execution.
Deploying a workflow transforms a one-off task into a **persistent, reusable process** that can be triggered on a schedule, by an external event, or called by Maia itself as part of a larger plan.
## Running Workflows Inside Sessions vs. Deployed
The interface remains the same whether you're actively working with Maia or reviewing automated runs. The difference is in how and when workflows execute.
### Running Inside the Session
**What it means:**
* You initiate the workflow directly in a session
* Watch execution in real-time
* Results delivered when complete
**Best for:**
* Testing new workflows
* One-off research tasks
* Tasks requiring human judgment
* Learning how Maia works
**Example:**
```
"Research the top 5 project management tools and
create a comparison doc"
```
You watch it run, review results, maybe tweak and re-run.
### Deployed Workflows
**What it means:**
* Workflow runs automatically based on one or more triggers
* No manual initiation required
* You can open any deployed workflow to see execution details
* Results saved in your workspace to access anytime
**Best for:**
* Recurring tasks (daily, weekly, monthly)
* Event-triggered actions (when X happens, do Y)
* Modular building blocks for Maia to use across plans
* Background data processing
* Scaling operations
**Example:**
```
Deploy the "competitor pricing check" workflow to run
every Monday at 9 AM and send results to #marketing-channel
```
Once deployed, the workflow runs automatically every week. You can open it anytime to see execution details and results.
## One Workflow, Multiple Triggers
Every deployed workflow can be triggered from multiple sources simultaneously. You assign **triggers** to control when and how a workflow runs:
Run on a schedule — hourly, daily, weekly, or monthly
Trigger via HTTP requests from external systems
Let Maia call the workflow as a reusable action within larger plans
A single workflow can have any combination of these triggers active at the same time. For example, a lead scoring workflow could run on a daily schedule, accept webhook calls from your website, and be available as an action Maia uses inside a larger lead generation pipeline — all simultaneously.
## Deploying a Workflow
There are two ways to deploy a workflow:
### Method 1: Deploy via Chat
Simply tell Maia you want to deploy your workflow, and Maia will suggest the appropriate trigger type with options.
**Example:**
```
"Deploy this workflow to run every Monday morning"
```
Maia will present an interactive suggestion in the chat with:
* Recommended trigger type (routine, webhook, or action)
* Pre-configured options based on your request
* A **Deploy** button to review and confirm settings
Click the deploy button in the chat suggestion to review the options and proceed with deployment.
### Method 2: Deploy via Workflow Board
Use the **Deploy workflow** button in the top-right corner of the workflow board.
**Steps:**
1. Click the **Deploy workflow** button in the top-right corner
2. Review the workflow details and click **Deploy** to activate
3. Once deployed, add triggers to control how the workflow runs:
* **Enable the Action trigger** to make the workflow available as a reusable action for Maia
* **Use the Add Trigger button** to attach routine (scheduled) or webhook triggers
4. Configure each trigger's settings and enable it
## Managing Triggers
Once deployed, you can manage all triggers from the deployed workflow's detail page:
* **Add triggers** — Attach additional routine or webhook triggers, or enable the action trigger
* **Enable/Disable** — Toggle individual triggers on or off without removing them
* **Edit** — Modify trigger settings (schedule, response configuration, etc.)
* **Delete** — Remove triggers you no longer need
* **Run manually** — Execute the workflow on demand at any time
## Workflow Limits by Plan
| Plan Tier | Active Workflows |
| -------------------- | ---------------- |
| **Free Plan** | 3 workflows |
| **Plus & Pro Plans** | Unlimited |
A "workflow" is a unique deployed workflow configuration. Adding multiple triggers to the same workflow still counts as **one** workflow toward your limit.
## Real-World Examples
### Example 1: Daily Competitor Monitoring
**Workflow:** Check 10 competitor websites for pricing changes
**Triggers:**
```yaml theme={null}
Routine: Daily at 8 AM (Mon-Fri)
Execution: Scrape pricing pages, compare to previous day
Delivery:
- If changes detected:
→ Email: Alert to pricing@company.com
- If no changes:
→ No notification (silent success)
```
**Benefit:** Team stays informed of market changes without manual checking
### Example 2: Weekly Lead Enrichment
**Workflow:** Process new leads from Google Form submissions
**Triggers:**
```yaml theme={null}
Routine: Every Monday at 9 AM
Webhook: Receives form submissions in real-time
Execution:
1. Read form responses from past week (routine) or process incoming lead (webhook)
2. For each lead, research company (web + LinkedIn)
3. Enrich with: industry, size, tech stack
4. Add to CRM with enriched data
Delivery:
- Update CRM records
- Email summary to sales@company.com
(# of new leads, top prospects)
```
**Benefit:** Sales team gets qualified, enriched leads without manual research
### Example 3: Monthly Financial Report
**Workflow:** Generate comprehensive financial report
**Triggers:**
```yaml theme={null}
Routine: 1st of every month at 6 AM
Execution:
1. Pull data from accounting software API
2. Calculate key metrics (revenue, expenses, profit, growth %)
3. Generate charts (revenue trend, expense breakdown)
4. Create 20-page PDF report with analysis
Delivery:
- Save to Drive: "Financial Reports/{{YYYY}}/{{Month}}.pdf"
- Email to: CFO, CEO, board members
```
**Benefit:** Consistent, professional monthly reporting with zero manual effort
## Execution Logs
Every deployed workflow maintains run logs:
### What's Logged
* **End time**: When execution completed
* **Duration**: How long it took
* **Steps completed**: Which actions ran successfully
* **Errors**: Any failures or issues
* **Results**: Summary of what was produced and what actions were taken
* **Trigger**: Which trigger initiated the run (routine, webhook, action, or manual)
Set up recurring workflows with flexible scheduling
# Routine Triggers
Source: https://docs.modularmind.app/workflows/routine-triggers
Run deployed workflows on a recurring schedule
## Turn Tasks into Routines
Routine triggers let you run deployed workflows at specific times without manual intervention — hourly, daily, weekly, or monthly. A single workflow can have multiple routine triggers, alongside webhook and action triggers.
## Setting Up a Routine Trigger
From your deployed workflow's detail page, click the **Add Trigger** button and select **Routine** to configure automatic execution at your preferred intervals.
### Scheduling Options
Triggers every hour at your chosen minute offset. Perfect for real-time monitoring tasks, frequent data updates, or time-sensitive automations that need to run throughout the day.
**Configuration:**
* Select minute offset (0–59 minutes past the hour)
Triggers once per day at your specified time. Ideal for daily reports, routine maintenance tasks, or regular content updates.
**Configuration:**
* Select the time of day when the workflow should execute
Triggers on specific days of the week at your chosen time. Great for weekly reports, periodic reviews, or tasks that align with your business week.
**Configuration:**
* Choose which day(s) of the week the workflow should run
* Select the hour to trigger execution
Triggers on specific day(s) of the month at your chosen time. Perfect for monthly reports, billing cycles, or end-of-month processes.
**Configuration:**
* Select which numbered day(s) of the month (1-31) to run
* Choose the hour to trigger execution
## Multiple Triggers on One Workflow
You can attach multiple routine triggers to a single deployed workflow. For example, a reporting workflow could have:
* A **daily** trigger for quick summary reports at 9 AM
* A **weekly** trigger for detailed analysis every Monday at 8 AM
* A **monthly** trigger for comprehensive reviews on the 1st
Each trigger runs independently. The workflow's execution logs show which trigger initiated each run.
## Use Cases
Generate and email performance reports every morning at 9 AM
Check competitor posts hourly and get notified of new content
Compile weekly analytics every Monday at 8 AM
Process billing and send invoices on the 1st of each month
## Best Practices
Match your schedule to your actual needs. Hourly schedules consume more resources — use them only when necessary.
Times are set based on your account's timezone. Ensure your chosen time aligns with your target audience or data availability.
Run your workflow manually first to verify it works correctly before setting up automatic execution.
Check your workflow runs regularly to ensure they're executing successfully and producing expected results.
You can pause or modify routine triggers at any time from your deployed workflow's settings without having to redeploy the entire workflow.
Learn how to trigger workflows from external events
# Webhook Triggers
Source: https://docs.modularmind.app/workflows/webhook-triggers
Trigger deployed workflows from external events and return intelligent responses
Use the [automating-with-maia](https://github.com/modularmindlab/maia-agent-skills) agent skill with your coding agent for a smooth end-to-end webhook integration setup — it includes reference scripts, response mode selection, and server-side implementation guidance.
## Event-Driven Automation
Webhook triggers turn deployed workflows into **intelligent APIs** that respond to external events — form submissions, payment notifications, chat messages, and more.
With webhook triggers, Maia becomes a **serverless backend** that processes inbound requests with AI-powered logic and returns structured responses.
## How Webhook Triggers Work
```mermaid theme={null}
graph LR
A[External System] -->|HTTP POST| B[Maia Webhook URL]
B --> C[Trigger Workflow]
C --> D[Process with AI]
D --> E[Stream Response]
E -->|SSE Events| A
```
### Basic Flow
From your deployed workflow's detail page, click the **Add Trigger** button and select **Webhook**.
Maia generates a unique URL:
```
https://hooks.modularmind.app/hooks/0nMMaZgKjmOirqosc8ohkeCuw9cV
```
Another system sends data to your webhook URL
```bash theme={null}
curl -X POST https://hooks.modularmind.app/hooks/0nMMaZgKjmOirqosc8ohkeCuw9cV \
-H "Content-Type: application/json" \
-d '{"email": "john@example.com", "message": "Interested in demo"}'
```
Your workflow receives the payload and executes
* Parse incoming data
* Run AI analysis, research, or generation
* Use actions
Maia streams back the result via SSE events
```
event: response
data: {"status":200,"contentType":"application/json","body":"{\"lead_score\":85,\"recommendation\":\"High priority - schedule demo ASAP\"}"}
event: done
data: {}
```
## Creating Webhook-Triggered Workflows
### Simple Example: Lead Enrichment
**Goal:** When a contact form is submitted, enrich the lead data before adding to CRM.
**Workflow:**
```
1. Receive form data (name, email, company)
2. Research company online (industry, size, funding)
3. Check LinkedIn for person's role
4. Calculate lead score (1-100)
5. Add to CRM with enriched data
6. Return lead score and priority
```
**Webhook URL:**
```
Maia provides: https://hooks.modularmind.app/hooks/L3wKjb7iEMFTwMJFZsroWsSoc731
```
**Usage:**
Your website form submits to this URL and receives an SSE stream. Once complete, the parsed response body contains:
```json theme={null}
{
"lead_id": "CRM_12345",
"score": 92,
"priority": "hot",
"reason": "Decision maker at Series B SaaS company"
}
```
## Inbound Payload: Flexible Schema
Unlike traditional APIs that require strict schemas, Maia webhooks are **intelligent**.
### Maia Understands Context
Send any data format, and Maia figures it out:
```json theme={null}
{
"customer_email": "jane@example.com",
"subject": "Need help with billing",
"message": "I was charged twice this month"
}
```
**Maia extracts:**
* Email: [jane@example.com](mailto:jane@example.com)
* Subject: Billing issue
* Problem: Double charge
```
email=jane@example.com&subject=billing&message=charged+twice
```
**Maia converts** to structured format automatically
```
Content-Type: multipart/form-data
file: invoice.pdf
description: "Please process this invoice"
```
**Maia receives** file, extracts data
```
Content-Type: multipart/form-data
file: screenshot.png
description: "Analyze this image"
```
**Maia detects and processes** JPEG/PNG images automatically
```
Content-Type: multipart/form-data
file1: contract.pdf
file2: receipt.jpg
file3: notes.txt
description: "Process all these documents"
```
**Maia handles** multiple files in a single request
```json theme={null}
{
"invoice": {
"filename": "invoice.pdf",
"content": "base64EncodedData...",
"encoding": "base64"
},
"supporting_docs": [
{
"filename": "receipt.jpg",
"data": "binaryData..."
}
]
}
```
**Maia processes** files embedded in JSON (base64 or raw binary)
```
Content-Type: multipart/form-data
file: documents.zip
instructions: "Extract and analyze all contents"
```
**Maia automatically extracts** ZIP files and processes contents (PDFs, images, text files)
```json theme={null}
{
"order_id": "12345",
"customer": {
"name": "John Doe",
"contact": {
"email": "john@example.com"
}
},
"items": [
{"product": "Widget", "qty": 5}
],
"attachments": {
"receipt": "base64Data..."
}
}
```
**Maia understands** arbitrary nested JSON structures
## Outbound Response: Controlling What Gets Returned
When adding a webhook trigger, you choose how Maia responds to incoming requests.
### Default Response (Immediate)
If you don't enable **"Respond to webhook with data"**, Maia returns immediately with one of these standard responses:
**Success:**
```json theme={null}
{
"type": "success",
"message": "Workflow execution triggered successfully."
}
```
**Error (Invalid Endpoint):**
```json theme={null}
{
"type": "error",
"code": "invalid_endpoint",
"message": "Invalid endpoint. Please check the path and try again."
}
```
**Error (Disabled Endpoint):**
```json theme={null}
{
"type": "error",
"code": "endpoint_disabled",
"message": "This endpoint is currently disabled."
}
```
The workflow executes asynchronously, but the caller doesn't wait for results.
### Custom Data Response (SSE Streaming)
Enable **"Respond to webhook with data"** to return workflow outputs back to the caller via **Server-Sent Events (SSE)**.
When enabled, you define **Body Parameters** — exactly like configuring [Custom Actions](/integrations/custom-actions). This tells Maia what data to include in the response.
Maia uses SSE streaming instead of a standard JSON response. This keeps the connection alive while the workflow executes, preventing timeouts on long-running workflows.
#### Connection Protocol
When a webhook trigger with response data enabled receives a request, the connection follows this lifecycle:
The server responds with `200 OK` and SSE headers:
```
Content-Type: text/event-stream
Cache-Control: no-cache
Connection: keep-alive
```
While the workflow executes, the server sends periodic heartbeat comments every 15 seconds to keep the connection alive:
```
: heartbeat
```
These are SSE comments and should be ignored by your client.
When the workflow completes, the server sends a `response` event containing the result:
```
event: response
data: {"status":200,"contentType":"application/json","body":"{...}"}
```
For text-based responses (`application/json`, `text/*`), the `body` field contains the raw string. For binary responses (images, PDFs, ZIPs), the `body` is base64-encoded and an `encoding: "base64"` field is included.
A `done` event signals the stream is complete:
```
event: done
data: {}
```
The connection closes after this event.
#### Error Handling
If the workflow fails to trigger, the server sends an `error` event instead:
```
event: error
data: {"type":"workflow_trigger_failed","message":"Failed to trigger workflow execution."}
```
#### Client Implementation Example
```javascript theme={null}
const eventSource = new EventSource('https://hooks.modularmind.app/hooks/your_hook_id', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ email: 'john@example.com' })
});
// Note: EventSource only supports GET requests.
// For POST requests, use fetch with a manual SSE parser:
const response = await fetch('https://hooks.modularmind.app/hooks/your_hook_id', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ email: 'john@example.com' })
});
const reader = response.body.getReader();
const decoder = new TextDecoder();
while (true) {
const { done, value } = await reader.read();
if (done) break;
const text = decoder.decode(value);
const lines = text.split('\n');
for (const line of lines) {
if (line.startsWith('event: ')) {
const eventName = line.slice(7);
// Next data: line contains the payload
}
if (line.startsWith('data: ')) {
const data = JSON.parse(line.slice(6));
// Handle event data
}
}
}
```
#### Supported Response Types
Maia can return:
* **Text data in JSON format** (structured responses) — delivered as-is in the `body` field
* **Images** (JPEG, PNG) — base64-encoded in the `body` field with `encoding: "base64"`
* **Text files** — delivered as-is in the `body` field
* **PDFs** — base64-encoded in the `body` field with `encoding: "base64"`
* **ZIP files** — base64-encoded in the `body` field with `encoding: "base64"`
Data, including files, can be generated throughout the workflow or downloaded from external sources.
#### Defining Response Parameters
For each parameter you want to return, specify:
**Name**: The field name in the response (e.g., `lead_score`, `generated_image`, `report_pdf`)
**Description**: Natural language explanation of what this field contains and where Maia should get the value from. This helps Maia understand which workflow outputs map to this response field.
**Type**: The type of data (text, file, etc.)
Response parameter definitions work identically to body parameters in [Custom Actions](/integrations/custom-actions#body-parameters). See that section for detailed parameter configuration guidance.
#### Example: Lead Scoring Response
**Configuration:**
```
Body Parameters:
- Name: lead_id
Description: The CRM record ID assigned to this lead
- Name: score
Description: Calculated lead score from 1-100 based on company research
- Name: priority
Description: Priority level (hot/warm/cold) based on the score
- Name: next_action
Description: Recommended follow-up action for the sales team
```
**SSE Stream:**
```
: heartbeat
: heartbeat
event: response
data: {"status":200,"contentType":"application/json","body":"{\"lead_id\":\"CRM_54321\",\"score\":88,\"priority\":\"hot\",\"next_action\":\"Schedule demo within 24 hours\"}"}
event: done
data: {}
```
**Parsed Response Body:**
```json theme={null}
{
"lead_id": "CRM_54321",
"score": 88,
"priority": "hot",
"next_action": "Schedule demo within 24 hours"
}
```
#### Example: Document Generation Response
**Configuration:**
```
Body Parameters:
- Name: report_pdf
Description: The generated PDF report file created by the workflow
Type: File (base64)
- Name: summary
Description: Brief text summary of the report contents
```
**SSE Stream:**
```
: heartbeat
: heartbeat
: heartbeat
event: response
data: {"status":200,"contentType":"application/json","body":"{\"report_pdf\":\"base64EncodedPdfData...\",\"summary\":\"Q4 2025 Sales Report - 23% growth, top performer: Enterprise segment\"}","encoding":"base64"}
event: done
data: {}
```
**Parsed Response Body:**
```json theme={null}
{
"report_pdf": "base64EncodedPdfData...",
"summary": "Q4 2025 Sales Report - 23% growth, top performer: Enterprise segment"
}
```
## Real-World Webhook Examples
### Example 1: Intelligent Chatbot Backend
**Setup:** Website chat widget → Maia webhook
**Workflow:**
1. Receive visitor message
2. Search knowledge base using custom action
3. If found: Generate helpful response
4. If not found: Create support ticket via custom action, send notification
5. Return response to chat widget
**Request from caller:**
```json theme={null}
{
"visitor_id": "vis_123",
"message": "What's your refund policy?",
"page_url": "/pricing"
}
```
**SSE stream received by caller:**
```
: heartbeat
event: response
data: {"status":200,"contentType":"application/json","body":"{\"response\":\"We offer a 30-day money-back guarantee on all plans. If you're not satisfied, contact support@company.com for a full refund.\",\"confidence\":0.92,\"escalate\":false,\"source\":\"Help Center: Refund Policy\"}"}
event: done
data: {}
```
**Parsed response body:**
```json theme={null}
{
"response": "We offer a 30-day money-back guarantee on all plans...",
"confidence": 0.92,
"escalate": false,
"source": "Help Center: Refund Policy"
}
```
**Result:** Instant, accurate responses powered by AI, with human fallback.
### Example 2: Invoice Processing
**Setup:** Email automation → Maia webhook
**Workflow:**
1. Receive invoice PDF as attachment
2. Extract data (vendor, amount, due date, line items)
3. Validate against purchase order using custom action
4. Add to accounting system via custom action
5. If > \$5000, create approval request via custom action
6. Return processing summary
**Request from caller:**
```json theme={null}
{
"from": "vendor@supplier.com",
"subject": "Invoice #12345",
"attachment_url": "https://storage.../invoice.pdf"
}
```
**SSE stream received by caller:**
```
: heartbeat
: heartbeat
event: response
data: {"status":200,"contentType":"application/json","body":"{\"status\":\"processed\",\"invoice_number\":\"12345\",\"vendor\":\"ABC Supplier\",\"amount\":3250.00,\"due_date\":\"2026-01-15\",\"accounting_id\":\"INV_789\",\"approval_required\":false,\"line_items\":[{\"description\":\"Widget A\",\"quantity\":100,\"price\":25.00},{\"description\":\"Widget B\",\"quantity\":50,\"price\":15.00}]}"}
event: done
data: {}
```
**Parsed response body:**
```json theme={null}
{
"status": "processed",
"invoice_number": "12345",
"vendor": "ABC Supplier",
"amount": 3250.00,
"due_date": "2026-01-15",
"accounting_id": "INV_789",
"approval_required": false,
"line_items": [
{"description": "Widget A", "quantity": 100, "price": 25.00},
{"description": "Widget B", "quantity": 50, "price": 15.00}
]
}
```
**Result:** Automated invoice processing with data extraction and validation.
### Example 3: Product Review Aggregator
**Setup:** Product review form → Maia webhook
**Workflow:**
1. Receive review text and rating
2. Analyze sentiment (positive/negative/neutral)
3. Extract key themes (quality, pricing, support, etc.)
4. Check for spam/fake reviews
5. If legitimate, publish to website via custom action
6. If mentions bug/issue, create support ticket via custom action
7. Return analysis
**Request from caller:**
```json theme={null}
{
"product_id": "prod_456",
"rating": 4,
"review": "Great product! Works as advertised. Customer
support was responsive when I had a question
about setup. Only issue: shipping took longer
than expected.",
"reviewer_email": "customer@example.com"
}
```
**SSE stream received by caller:**
```
: heartbeat
event: response
data: {"status":200,"contentType":"application/json","body":"{\"approved\":true,\"spam_score\":0.05,\"sentiment\":\"positive\",\"themes\":{\"product_quality\":\"positive\",\"customer_support\":\"positive\",\"shipping_speed\":\"negative\"},\"action_items\":[{\"type\":\"feedback\",\"category\":\"shipping\",\"message\":\"Customer experienced slow shipping\"}],\"published_url\":\"https://site.com/reviews/rev_123\"}"}
event: done
data: {}
```
**Parsed response body:**
```json theme={null}
{
"approved": true,
"spam_score": 0.05,
"sentiment": "positive",
"themes": {
"product_quality": "positive",
"customer_support": "positive",
"shipping_speed": "negative"
},
"action_items": [
{
"type": "feedback",
"category": "shipping",
"message": "Customer experienced slow shipping"
}
],
"published_url": "https://site.com/reviews/rev_123"
}
```
**Result:** Intelligent review moderation with actionable insights.
Learn how Maia protects your data and credentials