Use MCP Rating from Claude or ChatGPT
MCP Rating's MCP endpoint lets an AI assistant search our index while you chat: which servers fit a task, whether they started when we ran them, how they are maintained, and how to set them up. There is nothing to install and no account or key.
Endpoint: https://mcprating.io/mcp (Streamable HTTP, no authentication)
Claude
Customize › Connectors › + Add › Add custom connector. Enter the URL above and leave authentication empty. Free plans can add one custom connector.
ChatGPT
Settings › Security and login › Developer mode, then add the URL above. Whether Developer mode is available depends on your account and workspace.
Other MCP clients
Any client that supports remote MCP servers over Streamable HTTP can use the
URL above. To run MCP Rating locally instead, with sandboxed servers behind one
endpoint, use the gateway: npx -y @mcp-rating/gateway.
Tools
| Tool | What it does |
|---|---|
search_servers |
Search the index by query and category, sorted by quality, stars, downloads, recency or name |
get_server |
One server: our start test, maintenance, how to run it, the keys it declares, its tools, and the token cost of their definitions |
get_install_config |
Configuration for Claude Desktop, Cursor, or a generic client |
list_categories |
Categories and how many servers each has |
get_index_stats |
Index size, last crawl, top categories |
discover_server |
A server's live .well-known/mcp manifest, where it publishes one |
All the tools only read. None of them installs or runs anything on your machine.
Try asking
- "Find a well-maintained MCP server for PostgreSQL."
- "Did chrome-devtools-mcp start in MCP Rating's test, and what tools does it have?"
- "Give me the Claude Desktop config for firecrawl-mcp."
What the results mean
A start-test result means we launched the server with no credentials and it listed its tools. No result means we haven't tried it, not that it failed, and a server that needs an API key usually can't start without one. Quality scores are estimates from public signals (stars, downloads, documentation, maintenance), not security audits or endorsements. Server descriptions are written by each server's author. See the methodology.
Privacy
The endpoint receives the tool name and its arguments, such as a search phrase, and uses them only to answer. We count MCP searches but don't keep their text, and we never receive your conversation. Details are in the privacy policy.
Support
Questions, corrections or listing removal: [email protected].