feat(mcp): add Parallel Search MCP

[georgeatparallel]

Pull Request

Summary

• Add a source-backed Parallel Search MCP entry with anonymous setup, the web_search and web_fetch tools, current source links, and safety/privacy notes.

Submission Source

For direct content PRs:

• This PR changes exactly one content/<category>/<slug>.mdx file.
• The entry includes source/provenance URLs and practical install/use details.
• submittedBy and submittedByUrl match the PR author.
• I did not modify README.md, generated registry outputs, downloads, workflows, packages, scripts, or multiple content entries.
• I did not request HeyClaude-hosted /downloads/... package hosting for community-submitted ZIP/MCPB artifacts.
• This PR links the issue it resolves, or the no-issue rationale is written in Notes.

For platform/code/docs PRs:

• This is not a direct content submission.
• Changed routes/components/endpoints/tools are listed below.
• Screenshots or No visual impact are included when relevant.
• This PR links the issue it resolves, or the no-issue maintainer-lane rationale is written in Notes.

For registry API endpoint PRs:

• Route handler and central API contract changed together.
• Origin-check and rate-limit posture is stated below.
• OpenAPI/source generator impact is stated below; generated artifacts are not hand-edited unless maintainer/internal generation work is explicit.
• API contract tests were added or updated.
• Generator reproducibility was checked or the reason it was not checked is listed.

Schema and Quality Checks

• Content PR: pnpm validate:content:strict passed, or I am relying on CI.
• Platform/code PR: focused validation is listed below.
• Package artifact PR: pnpm validate:packages and pnpm scan:packages passed.
• No forbidden fields were added (viewCount, copyCount, popularityScore)
• Install/use/copy paths are practical and complete.
• Skill submissions include capability metadata when applicable (skillType, skillLevel, verificationStatus, verifiedAt, retrievalSources, testedPlatforms).

Quality Evidence

• Changed routes/components/endpoints/tools: One new source entry at content/mcp/parallel-search-mcp.mdx; it documents the web_search and web_fetch tools but does not change application code.
• Expected behavior: The registry can validate and publish a Parallel Search MCP detail entry with a copyable anonymous remote-MCP configuration.
• Important edge cases or invariants: The entry uses the current https://search.parallel.ai/mcp endpoint, keeps authentication optional, and does not edit generated artifacts.
• Backward compatibility notes: Content-only addition; no existing entry, schema, route, default, dependency, or generated output changes.
• Screenshots for frontend/page/UI changes:
  • Desktop: No visual impact; this uses the existing MCP entry renderer.
  • Mobile: No visual impact; this uses the existing MCP entry renderer.
• If screenshots do not apply, write No visual impact and explain why: No visual impact beyond rendering a new entry through the existing content template.
• Accessibility notes for UI changes: No UI component or interaction changes.
• Focused tests or reason tests are not practical: pnpm validate:content:strict and git diff --check passed.

Validation

• Direct content PR: I did not run generation or commit generated output.
• Platform/code PR: pnpm build passed, or the reason it was not run is listed below.
• I ran the focused checks listed above.
• I spot-checked the affected detail page(s), route(s), or integration surface(s), if applicable.

Notes

• Linked issue or no-issue rationale: No issue; the contribution is a focused, source-backed entry through the repository's documented direct-content PR path.
• Any caveats, follow-ups, or reviewer context: Current product documentation is authoritative for the endpoint because the linked source repository README still shows an older URL. The repository's normal CI can render and review the resulting detail page.

Disclosure: I work at Parallel.ai, which operates this MCP endpoint.