MCP Server

Connect Claude, Codex, Cursor, Gemini and other AI tools to your feedback workspace over the Model Context Protocol

The fdback.io MCP server lets AI tools — Claude, Claude Code, Codex, Cursor, Gemini, and any other Model Context Protocol client — read and manage your feedback workspace directly. Ask the model to triage feedback, draft a changelog, or summarize what users are asking for, and it works against your real data.

MCP access is a Pro feature and uses the same API keys as the REST API.

Server URL

https://app.fdback.io/api/mcp

This is a remote, Streamable HTTP MCP server — there is nothing to install. Authenticate with a live API key (create one under Settings → API) sent as a bearer token: Authorization: Bearer YOUR_API_KEY. The key acts as the user who created it, so its workspace role determines what the tools can do.

Connect your client

Pick your tool below, then replace YOUR_API_KEY with a live key.

Run this in your terminal — the workspace tools then appear in Claude Code:

claude mcp add --transport http fdback https://app.fdback.io/api/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Verify with claude mcp get fdback.

Claude Desktop only speaks stdio, so a remote server with a bearer token is reached through the mcp-remote proxy (requires Node.js). Add this to claude_desktop_config.json (Settings → Developer → Edit Config), then fully restart Claude:

{
  "mcpServers": {
    "fdback": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://app.fdback.io/api/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": { "AUTH_HEADER": "Bearer YOUR_API_KEY" }
    }
  }
}

The token's space lives in the AUTH_HEADER env var (note there is no space around the colon in the --header argument) — this avoids a known argument-escaping bug.

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json in your project, or use Settings → Tools & MCP:

{
  "mcpServers": {
    "fdback": {
      "url": "https://app.fdback.io/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Add to ~/.codex/config.toml (requires a recent Codex CLI), then start a new session:

[mcp_servers.fdback]
url = "https://app.fdback.io/api/mcp"
http_headers = { "Authorization" = "Bearer YOUR_API_KEY" }

Add to ~/.gemini/settings.json (user) or .gemini/settings.json (project). Use httpUrl (not url) so it connects over Streamable HTTP:

{
  "mcpServers": {
    "fdback": {
      "httpUrl": "https://app.fdback.io/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Config files are plaintext on disk. If you commit one (.mcp.json, .cursor/mcp.json, .gemini/settings.json), use an environment variable for the token instead of a literal key.

Available tools

Read

get_workspace — workspace metadata
list_filters — post types, statuses, and tags (the ids used for filtering)
list_feedback / get_feedback — list and read feedback posts
list_comments — a post's comment thread
list_changelogs / get_changelog — list and read changelog entries

Write

create_feedback / update_feedback — create and edit feedback (descriptions accept markdown)
create_comment — comment on a post or reply to a comment
set_vote / clear_vote — vote on a post
mark_shipped — mark a post as shipped

Changelog authoring

create_changelog / update_changelog — author changelog entries (content accepts markdown)
publish_changelog / unpublish_changelog — publish or revert to draft

Notes

Acts as the key owner. Every tool call runs as the user who created the API key and respects that user's workspace permissions.
Reads see internal data. Because the connection is authenticated, reads include internal and draft content, not just what's public.
No destructive tools. Deleting, archiving, and merging are intentionally not exposed over MCP.

Server URL

Connect your client

Available tools

Notes

On this page