Docs: add CLIRank as a free MCP example alongside Context7 in /codex/mcp
What
The Codex MCP docs page at https://developers.openai.com/codex/mcp uses Context7 as a canonical example of a free MCP server. Suggesting CLIRank as a complementary \"see also\" example - it covers the same shape (free, no auth, npx install) but for a different domain.
- Context7: free MCP for up-to-date library documentation
- CLIRank: free MCP for API discovery + comparison (which API to use, scored on agent-friendliness)
Both are useful at different points in an agent's flow: docs lookup vs choosing what to integrate in the first place.
Why this fits Codex specifically
A common Codex-driven task is \"add API X to my project\". Today, Codex's defaults often lean toward whatever was most popular in training data (Mailgun, Pinecone, SendGrid) — frequently not the best fit for headless agent use. CLIRank scores 416+ APIs on 8 agent-friendliness signals (SDK, env-var auth, headless support, JSON, CLI, curl docs, rate limits, machine-readable pricing) and exposes them as MCP tools so Codex can query at runtime instead of guessing.
Suggested addition to /codex/mcp
After the current Context7 install example, something like:
Another free MCP server worth knowing about is CLIRank - exposes a scored directory of 416+ APIs as MCP tools, useful when Codex needs to pick which API to integrate. Install: ``bash codex mcp add clirank -- npx clirank-mcp-server ``
Or, if a single example is preferred, swapping for CLIRank in the existing example would also work — the install pattern is identical.
Verification (60 seconds)
codex mcp add clirank -- npx clirank-mcp-server
Then in Codex: \"Use clirank to find the best transactional email API for an agent.\" Expects a tool call to search_apis, returns ranked results with scoring breakdowns. If Codex picks Resend or Postmark over Mailgun, the rubric is firing correctly.
Or via the hosted endpoint (no install):
[mcp_servers.clirank]
url = \"https://clirank-mcp.fly.dev/mcp\"
Disclosure
I built CLIRank. Free service, MIT-licensed MCP server, no monetisation. Submitting this because the pattern (scored API directory queryable at runtime) seems genuinely complementary to Context7 in the docs. Happy to PR a docs change directly if you point me at the source repo for developers.openai.com — couldn't find it via search.
Links
- Web: https://clirank.dev
- Methodology: https://clirank.dev/about
- MCP server source: https://github.com/alexanderclapp/clirank-mcp-server (MIT)
- Hosted endpoint: https://clirank-mcp.fly.dev/mcp
- Glama listing: https://glama.ai/mcp/servers/alexanderclapp/clirank-mcp-server (claimed, AAA quality)
- npm: https://www.npmjs.com/package/clirank-mcp-server
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗