# TimeDoctor MCP Server

> Read-only Model Context Protocol server exposing TimeDoctor workforce data —
> directory (users, projects, tasks, groups/tags, work schedules and leave) and
> time-tracking analytics (tracked/idle time, productivity, web & app usage,
> timeline, meetings) — to AI clients over remote Streamable HTTP. One bearer
> token = one workspace; every tool is read-only and tenant-scoped.

The authoritative, always-in-context guidance for an agent is the MCP server
`instructions` field returned at initialize. The pages below are the long-form
reference. For a single-fetch dump of the user-facing guides, see
[llms-full.txt](https://api2.timedoctor.com/api/2.0/mcp/llms-full.txt).

## Using the server
- [Agent usage guide](https://api2.timedoctor.com/api/2.0/mcp/guide/agent-guide): tools, parameters, pagination, rate limits, error meanings.
- [Adding tools (developers)](https://api2.timedoctor.com/api/2.0/mcp/guide/adding-tools): architecture and how to add a tool/domain.
- [Composable tools (TMCP-23)](https://api2.timedoctor.com/api/2.0/mcp/guide/composable-tools): multi-step drill-down aggregation and dispatch/gate rules.

## Getting started
- [Client setup](https://api2.timedoctor.com/api/2.0/mcp/guide/setup): per-client setup (Claude, ChatGPT, Gemini, Cursor, VS Code) + mcp-remote, auth headers, troubleshooting.

## Optional
- Endpoint: Streamable HTTP at /api/2.0/mcp/mcp
- Auth: Authorization: Bearer <token> + x-company-id: <company>
- Limits: ~10 req/s per credential, 500 rows/page, 256 KB/response, 8 s timeout.
- Full context: /api/2.0/mcp/llms-full.txt
- Discovery aliases: /api/2.0/mcp/.well-known/llms.txt, /api/2.0/mcp/.well-known/llms-full.txt