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

# Ship with MCP

> Connect AI tools to Emerge documentation and user data via MCP

Emerge provides two MCP (Model Context Protocol) servers for different purposes. Connect your AI tools to search documentation, query user data, or both.

## Two MCP Servers

<CardGroup cols={2}>
  <Card title="Docs MCP Server" icon="book" href="/ai/mcp-docs-setup">
    **Search and query this documentation**

    * Semantic documentation search
    * API reference lookup
    * Code example retrieval
    * Best for: Learning the API, troubleshooting
  </Card>

  <Card title="Query MCP Server" icon="database" href="/ai/mcp-query-setup">
    **Query consented user data via AI tools**

    * Search history, browsing, YouTube, ad interactions, receipts
    * Session-based credentials
    * Direct data access for AI agents
    * Best for: Building AI applications with user data
  </Card>
</CardGroup>

| Server        | Purpose              | URL                              | Auth Required         |
| ------------- | -------------------- | -------------------------------- | --------------------- |
| **Docs MCP**  | Search documentation | `https://docs.emergedata.ai/mcp` | No                    |
| **Query MCP** | Query user data      | `https://mcp.emergedata.ai/mcp`  | Yes (API token + uid) |

Both use the same Model Context Protocol and work with Claude, Cursor, VS Code, and Claude Code.

## Quick setup

<Tabs>
  <Tab title="Docs MCP">
    Add documentation search to your AI tool:

    ```bash theme={null}
    # Claude Code
    claude mcp add --transport http emerge https://docs.emergedata.ai/mcp
    ```

    ```json theme={null}
    // Cursor (.cursor/mcp.json)
    {
      "mcpServers": {
        "Emerge": {
          "type": "http",
          "url": "https://docs.emergedata.ai/mcp"
        }
      }
    }
    ```

    See [Docs MCP Setup](/ai/mcp-docs-setup) for all tools.
  </Tab>

  <Tab title="Query MCP">
    Add user data querying to your AI tool:

    ```bash theme={null}
    # Claude Code
    claude mcp add emerge-query "https://mcp.emergedata.ai/mcp?auth_token=YOUR_TOKEN&uid=YOUR_UID"
    ```

    ```json theme={null}
    // Cursor (.cursor/mcp.json)
    {
      "mcpServers": {
        "Emerge Query": {
          "url": "https://mcp.emergedata.ai/mcp?auth_token=YOUR_TOKEN&uid=YOUR_UID"
        }
      }
    }
    ```

    See [MCP Query Setup](/ai/mcp-query-setup) for all tools.
  </Tab>
</Tabs>

## What you can do

<CardGroup cols={2}>
  <Card title="Ask questions" icon="message-question">
    "How do I query user search history?" — AI searches docs and returns the answer with code examples
  </Card>

  <Card title="Generate code" icon="code">
    "Generate Python code to create a signed link" — AI finds SDK page and provides working implementation
  </Card>

  <Card title="Query user data" icon="database">
    "Get search history for the last week" — AI calls the Query MCP tool and returns results
  </Card>

  <Card title="Debug issues" icon="bug">
    "Why is my webhook signature invalid?" — AI finds troubleshooting info and suggests fixes
  </Card>
</CardGroup>

## How it works

```mermaid theme={null}
sequenceDiagram
    participant You
    participant AI Tool
    participant MCP Server
    participant Emerge

    You->>AI Tool: "How do I handle callbacks?"
    AI Tool->>MCP Server: Search docs
    MCP Server->>Emerge: Query content
    Emerge->>MCP Server: Return relevant pages
    MCP Server->>AI Tool: Structured context
    AI Tool->>You: Answer with code examples
```

## Benefits over web search

| Web Search                    | MCP Connection                  |
| ----------------------------- | ------------------------------- |
| May find outdated content     | Always current documentation    |
| Results from multiple sources | Single source of truth          |
| Generic code examples         | Emerge-specific implementations |
| Manual copy/paste             | Structured data for AI          |

## One-shot prompt for app builders

Use this prompt in tools like Lovable, Cursor, or Claude Code after connecting MCP:

```text theme={null}
You are building a full-stack app that integrates Emerge Link + Query.
Requirements:
- Use https://link.emergedata.ai/link/start to start consent.
- Generate HMAC-SHA256 signature over sorted raw params (exclude signature).
- Use signing_secret for HMAC; never expose secrets in the frontend.
- timestamp must be ISO 8601 (new Date().toISOString()).
- Generate random state, store it server-side, verify on callback.
- redirect_uri must be pre-approved in the Control Room.
- Never ask end-users for uid. If you send uid, use your internal user id. Otherwise store the callback uid.
- Query sync endpoints with uid: https://query.emergedata.ai/v1/sync/get_*
- Poll `GET /export/status/{uid}` and query only when the provider in `sources[]` is `data_ready: true`.
Return TypeScript and Python examples with async/await and error handling.
```

## Contextual menu

Every page in these docs includes a contextual menu with one-click AI integrations. Look for the menu icon on any page to:

* Copy as Markdown for pasting into AI chats
* Open directly in Claude or ChatGPT
* Copy the MCP server URL

Learn more in [Contextual Menu](/ai/contextual-menu).