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

# Connect Cursor and other MCP clients to Slaunt

> Connect Cursor or any Streamable HTTP MCP client to Slaunt. Without hooks, the model calls lifecycle tools itself.

Any Streamable HTTP MCP client with OAuth works. The difference from Claude Code and Codex is that there are no hooks, so Quaero only sees what the model reports. If the model never calls `run_begin`, the agent never appears. Install [agent instructions](https://docs.slaunt.ai/clients/agent-instructions.md) in the client's rules so it does.

```json .cursor/mcp.json theme={null}
{
  "mcpServers": {
    "slaunt": { "url": "https://central-mcp-ccr-production.up.railway.app/mcp" }
  }
}
```

Use `~/.cursor/mcp.json` for all projects. Other clients: point them at the same URL; OAuth is discovered from the `401` response.

A minimal turn from the model:

```text theme={null}
thread_begin  { request_id: "<uuid>" }       # once per chat
run_begin     { task: "Fix flaky login test" }
ccr_search    { query: "login test flake" }
run_step      { activity: "Patched retry logic", progress: 60 }
ccr_log_event { type: "change", summary: "Fixed flaky login test by awaiting session cookie." }
run_end       { outcome: "completed" }
```

If one connection carries several chats, tag each request with a [thread boundary](https://docs.slaunt.ai/protocol/transport.md#thread-boundaries), or they'll all share one thread.


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