MCP server for in-app feedback Triage feedback from your AI agent.

Set up in three steps

You need a Plus or Pro plan. The dashboard’s Connect a client tab shows these snippets with the URL and your new token already filled in.

Make a token

In the dashboard: Settings → MCP → Tokens → New token.

Name

So you can tell your tokens apart, e.g. “Claude Code on my laptop”.

Access

Read: see boards, feedback and comments. Read + write: also change status, and create, edit or delete feedback and comments.

Expires after

30, 90 (default) or 365 days.

The token is shown once, so copy it before you close the dialog. You can have up to 20.

Add the server to your client

Replace <your token> with yours.

claude mcp add --transport http fedo https://mcp.getfedo.com/mcp --header "Authorization: Bearer <your token>"

export FEDO_TOKEN="<your token>"

codex mcp add fedo --url https://mcp.getfedo.com/mcp --bearer-token-env-var FEDO_TOKEN

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "fedo": {
      "type": "remote",
      "url": "https://mcp.getfedo.com/mcp",
      "oauth": false,
      "headers": { "Authorization": "Bearer <your token>" }
    }
  }
}

{
  "mcpServers": {
    "fedo": {
      "httpUrl": "https://mcp.getfedo.com/mcp",
      "headers": { "Authorization": "Bearer <your token>" }
    }
  }
}

{
  "mcpServers": {
    "fedo": {
      "url": "https://mcp.getfedo.com/mcp",
      "headers": { "Authorization": "Bearer <your token>" }
    }
  }
}

{
  "servers": {
    "fedo": {
      "type": "http",
      "url": "https://mcp.getfedo.com/mcp",
      "headers": { "Authorization": "Bearer <your token>" }
    }
  }
}

{
  "mcpServers": {
    "fedo": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.getfedo.com/mcp", "--header", "Authorization:${FEDO_AUTH}"],
      "env": { "FEDO_AUTH": "Bearer <your token>" }
    }
  }
}

URL        https://mcp.getfedo.com/mcp
Transport  Streamable HTTP
Header     Authorization: Bearer <your token>

Not yet: claude.ai and ChatGPT. They need OAuth sign-in, which is coming soon.

Try a prompt

“Summarize this week’s feedback.” · “Which feedback in review has the most votes?” · “Mark the three most-voted planned items as in progress.” The last one needs a Read + write token.

Tools

The agent always acts as you, an admin on each of your boards, like the dashboard. One token works on every board you belong to. Statuses go over the wire as in_review, planned, in_progress, done (Shipped) and closed.

What it can’t do

It can’t vote, merge feedback, create or edit boards, manage board users, or act as one of your app’s users, whatever the token. The permissions table has the full list.

Example prompts

What the agent calls for each one. Feedback titles in these examples are illustrative. Five longer workflows, and how other feedback tools’ MCP servers compare, are in the guide Triage feature requests with an MCP server.

Permissions and safety

Feedback the agent creates includes your vote, as its author.

• Your yes before every write. Update and delete tools are marked destructive, so your client asks before it runs them. The server also tells the agent to show you what will change and wait for a yes before any write.
• Deletes by id only. delete_feedbacks and delete_comments take explicit ids, never a filter, and can’t be undone.
• Your users’ text is data. Feedback is written by your app’s users and could contain text aimed at the model. The server returns it only as data, and tells the model to treat it as data, never as instructions.
• No emails. Authors come back as a name and an admin flag. No email addresses.
• Tokens you can see and kill. Shown once when you make it. Expires after 30, 90 or 365 days. The list shows when each was last used; deleting one revokes it at once. Up to 20 per account.
• Bounded. Each write call takes 1 to 50 items and reports one result per item; a failed item doesn’t stop the others. 120 requests a minute per token.

FAQ

Which plans include the MCP server?

Plus and Pro. Starter, the free plan, doesn’t include it. Plus and Pro are in early access, so the price comes by email: see pricing.

Which AI clients work?

Claude Code, Codex, OpenCode and Gemini CLI in the terminal. Cursor and VS Code. Claude Desktop, through the mcp-remote bridge, which needs Node.js. And any other client that supports MCP over Streamable HTTP with a custom header.

Does it work with ChatGPT or claude.ai?

Not yet. Both need OAuth to connect, and fedo’s server takes a token in a header for now. They’re coming soon.

Can the agent delete things?

Only with a Read + write token, and only by id: the delete tools take explicit ids, never a filter like “everything in Closed”. Your client asks before running them, and deletes can’t be undone. A Read token can’t change anything.

What if a user writes instructions into their feedback?

The server returns feedback and comments only as data and tells the model to treat them as data, never as instructions. Writes still wait for your yes, and deletes still need explicit ids.

How do I revoke a token?

Delete it under Settings → MCP → Tokens. It stops working at once. Tokens also expire on their own after 30, 90 or 365 days.