MCP QUEEN

Give your AI agent an MCP research layer in one connection.

Add MCP Queen once and let ChatGPT, Claude, Claude Code, or your own agent find better-fit MCP servers before you connect them. Search by what you need done, compare what was actually observed, and keep the answer inside the same conversation.

Stop opening directory tabs, chasing stale repository links, and hand-comparing README claims. MCP Queen brings observed tools, live operational evidence, Trust Receipts, repository-mobility findings, and reviewed field reports into ChatGPT, Claude, or any compatible remote-MCP agent.

https://mcpqueen.com/mcp

What your agent gains

The simple version

Use it inside ChatGPT

Keep MCP research inside the chat where you are already planning, coding, or evaluating tools. Add MCP Queen as a custom MCP app, enable it for a conversation, and ask for candidates in ordinary language.

1. Open ChatGPT on the web and confirm your account or workspace currently supports custom MCP apps in developer mode.
2. Enable developer mode, create a custom app, and enter https://mcpqueen.com/mcp as the remote MCP endpoint.
3. Choose no authentication, scan the tools, and review the imported actions before creating or enabling the app.
4. Start a new chat, enable MCP Queen, and try one of the evidence prompts below.

Plan availability, workspace roles, and interface labels can change. Check OpenAI's current ChatGPT developer-mode guide. Developers using the API can follow the official remote MCP guide.

Use it inside Claude

Add MCP Queen as a custom connector and let Claude search the MCP ecosystem without leaving your conversation. Claude Code can use the same hosted endpoint from the terminal.

1. Open Customize → Connectors, select the add control, and choose Add custom connector.
2. Name it MCP Queen and enter https://mcpqueen.com/mcp.
3. Finish the setup, enable the connector for a conversation, and test a read-only discovery prompt.

Claude connects to remote custom connectors from Anthropic's cloud. Review Anthropic's current custom-connector instructions and security guidance.

claude mcp add --transport http mcpqueen https://mcpqueen.com/mcp

Put MCP discovery inside your own agent

Your agent should not need a separate scraping pipeline or hand-maintained MCP shortlist. Add the hosted endpoint as a remote MCP tool and it can search the Registry, inspect observed capabilities, and retrieve evidence during its existing workflow.

For an OpenAI Responses API agent, explicitly expose only the six read-only discovery and evidence tools. This keeps submit_feedback out of automated research runs.

{
  "type": "mcp",
  "server_label": "mcpqueen",
  "server_description": "Find MCP servers and inspect dated evidence before connecting.",
  "server_url": "https://mcpqueen.com/mcp",
  "allowed_tools": [
    "search_servers",
    "search_tools",
    "list_grades",
    "get_server_grade",
    "get_trust_receipt",
    "search_trust_evidence"
  ],
  "require_approval": "never"
}

Use approval requirements appropriate to your application. Do not send secrets or proprietary data in public MCP Queen search arguments.

Run the complete OpenAI Responses API example → · See framework integrations →

Start with one useful question

The payoff: MCP research becomes part of the work instead of a separate scavenger hunt. You keep the speed and context of your existing assistant while MCP Queen supplies a structured discovery and evidence layer.

Discovery without becoming the middleman

MCP Queen helps your assistant narrow the field and explain the evidence. It does not install every server, authorize one, or proxy the final connection.

Connection questions