Repository navigation
Conversation
clerk.com was listed only as an MCP server (via the Claude MCP directory feed). Clerk also ships a CLI and three documented HTTP APIs with published OpenAPI specs; the discovered entry for clerk.com carries them, but discovered surfaces are skipped for domains already present in the static catalog, so they never reached the listing. - sources/cli.json: the clerk CLI (npm i -g clerk / brew install clerk/stable/clerk, OAuth login via `clerk auth login`), docs at https://clerk.com/docs/cli - sources/openapi-manual.json: Backend API (2026-05-12), Frontend API (2026-05-12), and Platform API (beta) specs from the canonical clerk/openapi-specs repo Every fact verified against primary sources: the clerk/cli repo, the clerk/openapi-specs published YAMLs (titles, servers, security schemes), and clerk.com/docs. MCP endpoint re-probed (authless initialize 200). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
clerk.com now serves the newest spec of each API at a stable URL
(/docs/reference/spec/{bapi,fapi,platform}/latest.yml), so the catalog no
longer pins dated raw.githubusercontent files that go stale on each Clerk
API version release.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…dotsh into add-clerk-cli-and-apis # Conflicts: # sources/openapi-manual.json
mergeDeclared keyed existing surfaces by a single locator (spec || url || command || package || name), so an owner-declared surface carrying a spec could never match its discovered twin that only has a base url — clerk.com's re-discovery produced clerk-backend-api (discovered, no spec) alongside clerk-backend-api-2 (declared, spec), the duplicate (domain, type, name) shape validate:batch exists to catch. Matching now tries every facet a surface can be recognized by, ending with type+name — the same pair the batch validator treats as a duplicate. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
manovotny
marked this pull request as ready for review
September 23, 2026 15:43
Distinct APIs can share a base url — Clerk's Backend and Platform APIs both live at api.clerk.com/v1 — so matching a declared surface by url alone let the Platform declaration merge into the Backend surface and vanish. A surface with a spec now matches on the spec (and type+name), never its url; spec-less surfaces keep url matching. Regression test with both APIs added. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
clerk.com is listed only as an MCP server (it enters the catalog via the Claude MCP directory feed). Clerk also ships a CLI and three documented HTTP APIs with published OpenAPI specs; discovery finds them, but
buildDiscoveredEntriesskips discovered surfaces for domains already present in the static catalog, so the listing never picks them up.This adds them through the hand-maintained sources (same route as Railway's CLI entry):
sources/cli.json— theclerkCLInpm i -g clerk(also brewclerk/stable/clerk,npx clerk, andcurl -fsSL https://clerk.com/install | bash)clerk auth loginsources/openapi-manual.json— Clerk's three documented HTTP APIs, from stable vendor URLs that always resolve to the newest published spec (backed by clerk/openapi-specs):https://api.clerk.com/v1, secret-key bearer — spec: https://clerk.com/docs/reference/spec/bapi/latest.ymlsrc/lib/discover.ts— one pipeline fix, found by dogfooding the publishing flow: clerk.com now publishes/.well-known/integrations.json(validated against yourOwnerDeclaredDiscoveryschema),api-catalog, andmcp/server-card.json, and triggering re-discovery surfaced amergeDeclaredbug — it keyed existing surfaces by a single locator (spec || url || …), so a declared surface carrying a spec never matched its discovered twin that only has a base url. clerk.com's stored row now holdsclerk-backend-api(discovered) andclerk-backend-api-2(declared, spec) — the duplicate(domain, type, name)shapevalidate:batchexists to catch. Matching now tries every facet a surface can be recognized by, ending with type+name (the validator's own duplicate definition) — except that a spec-bearing surface never matches by url, since distinct APIs can share a base url (Clerk's Backend and Platform APIs both live atapi.clerk.com/v1). Tests added, including the shared-base-url case; a re-discovery of clerk.com after this deploys collapses the duplicates.Every fact verified against primary sources: the clerk/cli repo and docs, the published specs (titles,
servers, security schemes), and a live authless-initializeprobe ofhttps://mcp.clerk.com/mcp(200, streamable-http). Branch is merged up to current main (Datadog entries kept alongside in openapi-manual.json);bun test(55 pass) andbun run buildpass.Disclosure: I work at Clerk — the seed entries cover the static catalog side; the well-known files on clerk.com cover the discovery side.
🤖 Generated with Claude Code