Skip to main content
Project Echo’s MCP (Model Context Protocol) server lets AI tools query feedback and take actions—create requests, comment, vote, and more—using the same permissioned REST API under the hood. MCP is available on Free and Pro.

​
Before you start

  1. Open API & MCP in the Workspace nav.
  2. On the API tab, create a key with the permissions your agent should have.
  3. Copy the token (shown once) and key id.
  4. Open the MCP tab for a ready-to-copy client config.

​
Example MCP config

Add this to your MCP client (for example Cursor). Replace the placeholders with your key values:
Restart or reload the MCP client after saving.

​
If MCP fails to load

​
What you can do

Depending on key permissions, agents can read and update feedback data such as requests, comments, votes, notes, and changelog drafts—without leaving Claude, Cursor, or another MCP-compatible client. Publishing a changelog still happens in the Project Echo web app. Agents can also generate and fetch implementation prompts (Anthropic Claude Sonnet) with generate_implementation_prompt and get_implementation_prompt. See Generate implementation prompts with the API and MCP. For a reusable agent workflow, install the Project Echo agent skill. Weekly changelog drafts: Draft changelogs with an agent.

​
Security

  • Use least-privilege keys for agents.
  • Do not paste production tokens into shared chats or public repos.
  • Revoke and rotate keys from API & MCP if a token may have leaked.

​
Troubleshooting

  • Client cannot start the server — confirm Node/npx is available and the package name is @project-echo/mcp.
  • Auth errors — verify PE_API_KEY_ID and PE_API_TOKEN match an active key.
  • Missing tools/actions — widen permissions on a new key (tokens cannot change permissions after create).
Help: [email protected] · Help Center.