Copera MCP Server
The Copera MCP server is a hosted Model Context Protocol endpoint that lets AI clients — Claude Desktop, Claude Code, Cursor, and any other MCP-capable tool — read and write your Copera workspace directly. It exposes the same capabilities as the Copera Public API, packaged as native MCP tools an agent can discover and call.
You point your AI client at one URL, authenticate with a token, and the model gains a catalog of 37 tools spanning boards, documents, search, chat, notifications, drive, comments, and exports.
AI client (Claude / Cursor / …)
│ MCP over Streamable HTTP, Authorization: Bearer cp_pat_… or cp_oat_…
▼
Copera MCP server (https://mcp.copera.ai/mcp)
│ each tool call → HTTPS request to the Public API
▼
Copera Public API (https://api.copera.ai/public/v1)
How it relates to the Public API
The MCP server is a thin, stateless proxy over the Copera Public API. Every tool call becomes a typed HTTP request to https://api.copera.ai/public/v1/*, forwarding your token. The server keeps no session state and stores no data of its own — your token is validated by the Public API exactly as it would be for a direct REST call.
That means the MCP server inherits the Public API's permissions model, rate limits, and behavior. If you can do something with a Personal Access Token over REST, you can do it through MCP — and an AI client gets it as a tool it already knows how to use.
Parity and current gaps
The MCP tool catalog tracks the Public API closely, but it is not a one-to-one mapping of every endpoint. A few Public API capabilities have no MCP tool today:
- Row column-level markdown via a dedicated endpoint beyond the
set_row_markdown/get_row_markdownsurface. - Row authentication (the table-row authenticate endpoint).
- Drive multipart upload (presigned multi-part upload start/finalize) — the MCP drive tools read and create folders but do not upload file bytes.
- Bookings — no MCP tools are exposed for the Bookings surface.
For the full capability matrix across the Public API, MCP, CLI, and Copera AI tools, see the parity overview.
Which AI clients can use it
Any client that speaks the MCP Streamable HTTP transport can connect, including:
- Claude Desktop and Claude Code (via a remote MCP server / connector)
- Cursor
- The MCP Inspector for manual testing and debugging
- Any custom agent built on an MCP SDK
Next steps
Endpoint URL, client config, and testing with the MCP Inspector.
Personal Access Tokens vs. OAuth, the OAuth flow, and how scopes map to tools.
All 37 tools grouped by domain, with the Public API capability each maps to.
Concrete agent workflows and example tool-call sequences.