The ConnectorReport MCP server

Let your agent check whether an MCP connector actually works before it tries one. This server answers from the same hands-on test data as this site (35 connectors tested so far). It is free, public and read-only. No sign-in or API key.

Endpoint: https://connectorreport.com/api/mcp (Streamable HTTP)

Install

Claude

Claude.ai or Claude Desktop: Settings, Connectors, Add custom connector, then paste the endpoint URL. Claude Code:

claude mcp add --transport http connectorreport https://connectorreport.com/api/mcp

Cursor

Add this to ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "connectorreport": { "url": "https://connectorreport.com/api/mcp" }
  }
}

Any other client

Clients that speak Streamable HTTP just need the URL. For stdio-only clients, bridge with mcp-remote:

{
  "mcpServers": {
    "connectorreport": { "command": "npx", "args": ["-y", "mcp-remote", "https://connectorreport.com/api/mcp"] }
  }
}

Tools

Tool	What it returns
`list_connectors`	Every tested connector with status, pass rate and last test date. Optional `category` filter.
`get_connector_report`	Full report for one slug: status, pass rate, tested date, each task, gotchas, fix recipes and page URL.
`check_connector_status`	Current Connector Weather (Clear, Cloudy or Stormy) for a name, plus the last test date.
`compare_connectors`	Side-by-side results for a list of slugs.
`search_reports`	Search tasks, notes and gotchas, for example "OAuth" or "token".

Try asking your agent

"Before we connect GitHub, check ConnectorReport for known problems."
"Which docs MCP servers passed every test?"
"Compare Context7 and DeepWiki."

Fair use

The server is lightly rate limited to about 60 requests per minute per IP. Results come only from real test runs, and each one carries its test date. See Connector Weather for the same data in a browser.