Field-by-field comparison

Compare the evidence before choosing an MCP.

Each signal stays separate, missing facts remain visible, and the column order follows your selection.

MCPs in this comparison

Uncheck an MCP to remove it. Comparing two or three keeps each signal in its own row.

Open canonical comparison link

Comparable alternatives

Shown only when capability or industry context overlaps.

Side-by-side MCP evidence comparison
Evidence fieldio.github.blazickjp/arxiv-mcp-serverWatchlist candidateContext7 MCPMCP100 Selected
Task fit
Best for
  • Searching arXiv and retrieving or organizing papers through an MCP research workflow
  • Current library and API documentation inside coding agents
  • Read-only technical lookup with a small two-tool surface
  • Users who want remote or local npx connection options
Not ideal for
  • Systematic literature reviews, paywalled coverage, citation-complete searches, or judging research quality from metadata alone
  • Private internal documentation not indexed by Context7
  • Guaranteed authoritative answers without checking the cited upstream docs
  • Offline-only development
Avoid when
  • A downstream answer would cite or summarize a paper without checking its identity, version, content, and publication status
  • Your policy prohibits sending formulated documentation queries to a hosted service
  • You require complete or guaranteed-correct community-contributed documentation
  • You cannot keep an API key out of shared configuration
Provenance
Provenance details

Community implementation

io.github.blazickjp · Publisher source ↗

First-party MCP

Upstash · Publisher source ↗

Maintenance
Maintenance details

Repo: Aug 14, 2026

Package: Jul 29, 2026

Repo: Aug 14, 2026

Package: Aug 11, 2026

Popularity evidence
GitHub stars

3,052

GitHub stars · checked 2026-08-14T18:46:26.000Z

60,740

GitHub stars · checked 2026-08-14T18:46:26.000Z

30-day package downloads

79,761

PyPI Stats package downloads · checked 2026-08-14T18:46:26.000Z

3,916,841

30-day package downloads · checked 2026-08-14T18:46:26.000Z

External adoption evidence
  • github: repository stars: 3,052Exact attributable signal.github: repository stars · Admission snapshot · checked 2026-08-14T18:46:26.000ZEvidence source ↗
  • PyPI Stats: package downloads: 79,761Non-comparable contextual signal.PyPI Stats: package downloads · Provider last month · checked 2026-08-14T18:46:26.000ZEvidence source ↗
  • github: repository stars: 60,740Exact attributable signal.github: repository stars · Admission snapshot · checked 2026-08-14T18:46:26.000ZEvidence source ↗
Client coverage
Client coverage details
Claude CodeLocal stdio
Claude DesktopLocal stdio
CodexLocal stdio
VS CodeLocal stdio
CursorLocal stdio
OpenCodeLocal stdio
Claude CodeStreamable HTTP
Claude DesktopStreamable HTTP
CodexStreamable HTTP
VS CodeStreamable HTTP
CursorStreamable HTTP
OpenCodeStreamable HTTP
Client coverage noteLocal stdio runs on your computer. Streamable HTTP connects to a remotely hosted MCP server.
Access and data
AuthenticationNo authentication is documented for local stdio. HTTP defaults to 127.0.0.1; the publisher requires authentication and network controls if a reverse proxy exposes it.Remote MCP accepts CONTEXT7_API_KEY as an HTTP header. The local package accepts --api-key. The CLI can use OAuth to generate an API key; unauthenticated documentation commands may have lower limits.
CostThe Apache-2.0 project lists no MCP fee. External service quotas or provider charges are not documented.A free API key and free usage are documented, with higher-rate or team/enterprise offerings potentially requiring a paid plan. Applicable limits should be checked in the current Context7 dashboard.
PermissionsSearches arXiv, retrieves abstracts and LaTeX, downloads and converts papers, reads local papers, queries citations, exports citations, and creates or updates local topic watches and indexes.The MCP surface resolves library identifiers and queries documentation. It does not require repository or filesystem write access for the documented hosted mode.
Data handlingPapers and indexes stay local while searches, source retrieval, citation graphs, and downloads call external services. Paper text and LaTeX are untrusted external content; retention and model-training behavior are not documented.Context7 states that formulated queries are sent to its API while full prompts, source code, and conversation history are not transmitted; its tool descriptions instruct clients to remove secrets, personal data, and proprietary code.
Limitations
Tradeoffs
  • Natural-language paper discovery is convenient, while arXiv coverage, metadata, versions, and model-generated summaries can be incomplete or wrong.
  • An API key is recommended for higher limits, adding account and secret management.
  • Context7 says it sends formulated queries rather than full prompts or source code, but queries still leave the client.
  • Community-contributed projects can be incomplete or inaccurate and should be checked against primary documentation.
Risk contextDownloaded papers can contain prompt-injection instructions, and exposed HTTP could reveal research tools. Keep HTTP private, use approval controls, and treat paper content as untrusted.Primary risks are query disclosure, stale or inaccurate indexed documentation, and accidental API-key exposure. Keep queries free of proprietary details and verify consequential advice against first-party library documentation.
Evidence date
Editorial review2026-08-112026-07-24
Candidate evidence2026-08-14T18:46:26.000Z2026-08-14T18:46:26.000Z

Popularity, maintenance, fit, permissions, and client support are independent evidence fields.

Read the evidence method →