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

# MCP — Judit Model Context Protocol

> Connect AI assistants (Claude, GPT, etc.) directly to the Judit API via MCP (Model Context Protocol). Reach lawsuit queries, tracking and registration data from any agent.

The Judit documentation ships with a Model Context Protocol (MCP) server. Connect any AI client and ask questions about the API in natural language — the assistant searches the docs, cites the exact pages, and helps you build with Judit faster, without copy-pasting from the browser.

## Chat now (no install)

Click below to open Claude or ChatGPT with the full Judit documentation preloaded as context. Best for one-off questions or quick exploration.

<CardGroup cols={2}>
  <Card title="Chat with Claude" icon="comment" href="https://claude.ai/new?q=I%27m%20integrating%20with%20the%20Judit%20API%20%28Brazilian%20legal%20data%29.%20Please%20read%20https%3A%2F%2Fdocs.judit.io%2Fllms-full.txt%20for%20the%20full%20context%2C%20then%20ask%20me%20what%20I%20want%20to%20build.">
    Opens Claude.ai with the full docs preloaded. Zero install.
  </Card>

  <Card title="Chat with ChatGPT" icon="comments" href="https://chatgpt.com/?q=I%27m%20integrating%20with%20the%20Judit%20API%20%28Brazilian%20legal%20data%29.%20Please%20read%20https%3A%2F%2Fdocs.judit.io%2Fllms-full.txt%20for%20the%20full%20context%2C%20then%20ask%20me%20what%20I%20want%20to%20build.">
    Opens ChatGPT with the full docs preloaded. Zero install.
  </Card>
</CardGroup>

<Tip>
  Both buttons reference [`https://docs.judit.io/llms-full.txt`](https://docs.judit.io/llms-full.txt), an autogenerated single-file dump of the entire documentation, optimized for AI ingestion in one fetch.
</Tip>

## Install the MCP server

For repeated use across Claude Desktop, Cursor, Claude Code, ChatGPT, or any other MCP-compatible client, install the MCP server once and the tools become available natively in your chat:

<Card title="MCP server URL" icon="link" href="https://docs.judit.io/mcp">
  `https://docs.judit.io/mcp`
</Card>

<Tabs>
  <Tab title="Claude Desktop">
    Open Claude Desktop, go to **Settings → Developer → Edit Config**, and add:

    ```json theme={null}
    {
      "mcpServers": {
        "judit": {
          "url": "https://docs.judit.io/mcp"
        }
      }
    }
    ```

    Restart Claude Desktop. The **judit** tools appear in the chat input.
  </Tab>

  <Tab title="Cursor">
    Open the IDE's **Settings → MCP** (or edit `~/.cursor/mcp.json`) and add:

    ```json theme={null}
    {
      "mcpServers": {
        "judit": {
          "url": "https://docs.judit.io/mcp"
        }
      }
    }
    ```
  </Tab>

  <Tab title="Claude Code">
    From your terminal:

    ```bash theme={null}
    claude mcp add --transport http judit https://docs.judit.io/mcp
    ```
  </Tab>

  <Tab title="ChatGPT">
    In ChatGPT, open **Settings → Connectors → Add custom connector** and paste:

    ```
    https://docs.judit.io/mcp
    ```

    Available on plans that support MCP connectors (Plus, Pro, Team, Enterprise).
  </Tab>

  <Tab title="Other clients">
    Any MCP client that speaks HTTP transport works — point it at `https://docs.judit.io/mcp`.

    For clients that only speak stdio (older configs), wrap the URL with `mcp-remote`:

    ```json theme={null}
    {
      "mcpServers": {
        "judit": {
          "command": "npx",
          "args": ["-y", "mcp-remote", "https://docs.judit.io/mcp"]
        }
      }
    }
    ```
  </Tab>
</Tabs>

## What it does

Once connected (via either path above), the AI gets two tools it uses automatically when you ask anything Judit-related:

| Tool                               | Purpose                                                                                                          |
| :--------------------------------- | :--------------------------------------------------------------------------------------------------------------- |
| `search_judit_docs`                | Full-text search across every page of this documentation, returning the most relevant excerpts with their paths. |
| `query_docs_filesystem_judit_docs` | Read any specific doc page by path (`cat`, `head`, `rg`, etc.) when the AI needs the full context.               |

You don't call them yourself — just chat naturally and the model decides when to use them.

## Example prompts

* *"How do I create a tracking by CNJ? Show me the exact request body."*
* *"What's the difference between Hot Storage and a regular request?"*
* *"Write a Python snippet that creates a request, polls until it's done, and prints the lawsuits."*
* *"Which filters can I pass to `subject_codes`? Cite the page."*
* *"Walk me through the criminal execution flow end to end."*

The model will reference the exact docs pages it used, so you can verify the answer.

## What it does not do

This MCP **reads documentation only**. It does not call the Judit API on your behalf — no real lawsuits are queried, no requests are created, no costs are incurred.

To actually hit the API, use one of these:

<CardGroup cols={2}>
  <Card title="Postman Collection" icon="paper-plane" href="/en/resource/postman-collection">
    Run live requests with prefilled bodies and auto-captured `request_id` chaining.
  </Card>

  <Card title="REST API" icon="code" href="/en/introduction/authentication">
    Call the endpoints directly from your code with your `api-key`.
  </Card>
</CardGroup>

<Note>
  The MCP server is hosted by our documentation platform and stays in sync with this site automatically. Every doc update is reflected in the MCP responses on the next call — no version pinning, no redeploy needed on your side.
</Note>