> ## Documentation Index
> Fetch the complete documentation index at: https://docs.babou.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server Overview

> The agentic interface to the Babou engine

## What MCP gives you

The Babou MCP server is the agentic interface to the same engine the [REST API](/api-reference/overview) exposes. Same product. Different consumer. Instead of your application code making calls, an AI assistant (Claude, Cursor, anything MCP-aware) makes them on the user's behalf.

<Info>
  **MCP or REST?** Use MCP when an AI assistant is the driver: agents picking templates, assembling chapters, watching for product changes and reacting. Use the REST API for direct programmatic control from your own application.
</Info>

## What changes with MCP

<CardGroup cols={2}>
  <Card title="Agentic, not chat" icon="robot">
    The agent picks the project shape, opens chapters, fills them with prompts, exports the result. You describe outcomes, not API calls.
  </Card>

  <Card title="Watches your product" icon="eye">
    Point an agent at a release note, a Figma frame, or a deployed page, and it builds the matching marketing video.
  </Card>

  <Card title="Picks the right building block" icon="grid-2">
    Agents can pick a Template, drop in Blocks, or invoke an App for the input it has, instead of building from scratch every time.
  </Card>

  <Card title="Async-aware" icon="clock">
    Tools that are slow (prompts, exports) wait for completion or return a clean status the agent can poll.
  </Card>
</CardGroup>

## Available tools

The Babou MCP server exposes six tools today:

### Projects

<AccordionGroup>
  <Accordion title="CreateProject" icon="plus">
    Create a new project with a name and description.
  </Accordion>

  <Accordion title="ListProjects" icon="list">
    List all your projects.
  </Accordion>
</AccordionGroup>

### Chapters

<AccordionGroup>
  <Accordion title="AddChapter" icon="file-plus">
    Add a chapter to a project with a name and optional duration.
  </Accordion>

  <Accordion title="ListChapters" icon="list-ol">
    List all chapters in a project, with their current status.
  </Accordion>
</AccordionGroup>

### Prompts

<AccordionGroup>
  <Accordion title="AddPrompt" icon="message-pen">
    Submit a prompt to a chapter, describing what to build. Waits up to 30 seconds for completion.
  </Accordion>
</AccordionGroup>

### Export

<AccordionGroup>
  <Accordion title="ExportProject" icon="file-export">
    Export a completed project to a downloadable video file.
  </Accordion>
</AccordionGroup>

## How it lands

<Steps>
  <Step title="Set up the server">
    Drop the Babou MCP server into your AI assistant's config. See the [setup guide](/mcp/setup).
  </Step>

  <Step title="Authenticate">
    Connect with your Babou API key or OAuth.
  </Step>

  <Step title="Point the agent at a real surface">
    A release note, a Figma frame, a deployed page, a screenshot. The agent reads it.
  </Step>

  <Step title="Agent runs the pipeline">
    The agent uses the tools above to assemble the project, fill chapters, and export.
  </Step>

  <Step title="Get the output">
    Download the rendered video, or iterate.
  </Step>
</Steps>

## MCP vs REST

|                          | MCP server                                              | REST API                                            |
| ------------------------ | ------------------------------------------------------- | --------------------------------------------------- |
| **Best for**             | Agents driving the work                                 | Application code driving the work                   |
| **Caller**               | Claude, Cursor, MCP-aware tools                         | Your services                                       |
| **Style**                | Outcomes ("build the launch ad from this release note") | Endpoints (`POST /projects`, `POST /chapters`, ...) |
| **Auth**                 | OAuth or API key                                        | API key                                             |
| **Async**                | Tools wait for completion or return pollable status     | Manual polling                                      |
| **When to reach for it** | Exploration, iteration, hands-off pipelines             | Production integrations, scheduled jobs             |

<Note>
  The MCP server is a thin wrapper over the REST API. Anything an agent can do via MCP, your code can do via REST, and vice versa.
</Note>

## Getting started

<CardGroup cols={2}>
  <Card title="Setup guide" icon="gear" href="/mcp/setup">
    Install and configure the MCP server
  </Card>

  <Card title="Tool reference" icon="wrench" href="/mcp/tools/create-project">
    Every tool, with parameters and examples
  </Card>

  <Card title="Examples" icon="code" href="/mcp/examples">
    Real flows: launch ad, Figma demo, batch variants
  </Card>

  <Card title="REST API" icon="webhook" href="/api-reference/overview">
    Drive Babou from your own application code
  </Card>
</CardGroup>

## Supported clients

The Babou MCP server works with any MCP-compatible client:

* **Claude Desktop** (Anthropic)
* **Claude for VS Code**
* Cursor
* Any custom MCP client

<Tip>
  Claude Desktop is the simplest place to start. Follow the [setup guide](/mcp/setup).
</Tip>