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

# MCP server

> Use every Mindcase API from Claude, ChatGPT, Cursor and other MCP clients.

If your client speaks the Model Context Protocol, add Mindcase as a remote MCP server. It runs the same APIs and bills the same prepaid wallet as the REST API.

* **Server URL:** `https://mcp.mindcase.co/` (Streamable HTTP)
* **Auth, option 1: OAuth.** Sign in with your Mindcase account. Clients that support Client ID Metadata Documents (Claude.ai, ChatGPT, Vercel Connect) need no client ID or registration.
* **Auth, option 2: API key.** Send `Authorization: Bearer mk_live_...` on every request.
* **What's included:** every Mindcase API except the AI media generators (`ai/images`, `ai/videos`, `ai/speech`), which you can still call through the [REST API](https://docs.mindcase.co/running-apis.md).

## Setup

> [tab] Claude.ai
>
> Settings → Connectors → Add custom connector → paste `https://mcp.mindcase.co/` → Connect, then sign in to Mindcase. No key needed.

> [tab] Claude Code
>
> ```bash theme={null}
> claude mcp add --transport http mindcase https://mcp.mindcase.co/ \
>   --header "Authorization: Bearer mk_live_xxxxxxxxxxxxxxxxxxxx"
> ```

> [tab] Cursor and others
>
> In `~/.cursor/mcp.json`, or the same shape in most other clients:
>
> ```json theme={null}
> {
>   "mcpServers": {
>     "mindcase": {
>       "url": "https://mcp.mindcase.co/",
>       "headers": { "Authorization": "Bearer mk_live_xxxxxxxxxxxxxxxxxxxx" }
>     }
>   }
> }
> ```

## Tools

| Tool | What it does |
| - | - |
| `list_agents()` | The catalog: every API's `group`, `slug`, name, description, unit and price per row |
| `get_agent_schema(group, slug)` | The API's inputs, input modes and a ready-to-run example |
| `run_agent(group, slug, params)` | Validates `params` first (a bad call returns the schema and an example, and nothing is charged), then runs the API, waits up to about 2 minutes, and returns the first 100 rows with `job_id`, `row_count` and `truncated` |
| `fetch_results(job_id, offset, limit)` | Pages through a finished job's rows, or picks up a job `run_agent` stopped waiting on |

The usual order is `list_agents` → `get_agent_schema` → `run_agent`, then `fetch_results` if `truncated` is true.

## For coding agents

Drop [mindcase.co/skills.md](https://mindcase.co/skills.md) into your project and agents like Claude Code and Cursor will know every API, its inputs and its outputs without fetching anything.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.