Head to head · Documentation lookup · October 2026 research run

GitMCP vs Ref

Ref scores 49.4 (D) on agent readiness against GitMCP's 40.2 (E), and leads in 5 of 7 scored categories. GitMCP leads on payments & pricing. Both do documentation lookup.

Best code, repository and documentation tools for AI agents · All 62 code comparisons

Which one, for what

GitMCP E

Best for Quick, keyless library docs for public GitHub projects in a coding agent.

Ahead on

  • Payments & pricing, 60 against 35

Also in its favour

  • No key needed to call it
  • Open source

Watch for

No commit, release or maintainer reply on the repository since 8 May 2026

Ref D

Best for Coding agents that need current library documentation plus search over a team's private repositories and PDFs through two small tools.

Ahead on

  • Schema & documentation, 68 against 56
  • Agent ergonomics, 67 against 58
  • Security & auth, 63 against 42
  • Maintenance & community, 35 against 8
  • Transparency & trust, 69 against 36

Also in its favour

  • Runs on your own machine

Watch for

No npm release since 3.0.3 on 4 November 2025

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Transparency & trust Ref +33

GitMCP 36

Apache-2.0, with the whole service in the public repository (30). The README says the service stores no personal data or queries and keeps no repository content permanently, but there is no privacy policy or terms page (8). No deprecation policy or dated notices; SSE was removed on 8 May 2026 while the README still shows an SSE config for VS Code (0). The hosting stack (Cloudflare Workers, Workers AI, Vectorize, an Analytics Engine dataset) is visible in wrangler.jsonc but no processor is named for users, and invocation logs were turned off on 8 May (5). 30 + 8 + 0 + 5 = 43.

Ref 69

Per the 8 October check, not re-read on 4 November (ref.tools and docs.ref.tools are paused for this run pending a ruling on Ref's terms). The stdio server is MIT, and it is a thin client. The search service is closed. Its terms and privacy policy are drawn by script and were read on 8 October only from the script files the pages load, which this run doesn't count, so both are unread here. The line is held, as Prefactor's licence and data-handling lines were when its script-drawn terms were struck (18). The security page describes encryption, tenant isolation, zero-data-retention terms at model providers and a seven-day limit on agent transcripts. No DPA was found, and Centure.ai is named on the comparison page but not in the subprocessor table (18). No deprecation policy found. The pricing page says legacy $9 and $19 subscriptions continue unchanged (3). 19 subprocessors are named with what each can see. Data locations aren't stated (14). 18+18+3+14 = 53.

Maintenance & community Ref +27

GitMCP 8

The last commit on main is 8 May 2026, 182 days before this check, and there are no tags or releases (0). Nothing in the last 90 days (0). Since May, outside issues and pull requests (#244 to #271, among them the security report #249 and fixes #256, #267 and #250) have gone without a maintainer reply or merge (3). Not found in the official MCP registry under any namespace (0). A run-tests workflow with vitest and a Playwright end-to-end workflow exist on push and pull request (5). 0 + 0 + 3 + 0 + 5 = 8.

Ref 35

No dated release or API change of Ref Context is readable within 180 days. The ref-tools-mcp package was last published on 4 November 2025 (3.0.3, the version-bump commit is dated the same day in the repository), 365 days before this check. The 31 July and 30 September 2026 updates were read on 8 October from mailchi.mp, whose robots.txt now disallows every crawler, so they're struck; the docs changelog index gives months only, and both covered Ref Plans. The repository's last commits, on 28 September 2026, are dependency updates, which can't set recency (0 of 30). No releases or dated entries in the 90 days from 6 August (0). Per the 8 October reading of GitHub's API, closed issues show two to five comments each and September's open pull requests one comment each, and the docs promise a reply within one business day (15). The official MCP registry lists tools.ref/ref-tools-mcp under Ref's own domain, at 3.0.1 from 28 October 2025, still on 4 November (15). Dependabot alerts were cleared on 28 September 2026. The only workflow builds a bundle on release, with no tests (5). 0+0+15+15+5 = 35.

Payments & pricing GitMCP +25

GitMCP 60

Nothing to buy, so no payment protocol (0). Free, stated in the README FAQ (20). No account or card (20). An agent connects with a URL and no signup or key (20). 0 + 20 + 20 + 20 = 60.

Ref 35

Per the 8 October check, not re-read on 4 November (ref.tools and docs.ref.tools are paused for this run pending a ruling on Ref's terms). No x402, MPP or L402 (0). One credit per tool call and $10 per 1,000 extra credits are published without a login (20). The free plan is 200 one-time credits that never expire, not a recurring allowance, and the docs don't say whether sign-up needs a card (15). A person has to sign up in a browser to get a key (0).

Security & auth Ref +21

GitMCP 42

No credentials at all, so nothing to leak or scope (20). Every tool is a read of public content (15). Repository docs, code and fetched URLs reach the model unmarked, and fetch_generic_url_content fetches any absolute URL a document mentions, with no prompt-injection guidance in the README (2). No per-call visibility for the caller (0). SECURITY.md routes reports to a Google Form, but the server-side request forgery report in issue #249 (12 June 2026) and its proposed fix #256 have had no maintainer reply in almost five months. No certifications or bounty (5). 20 + 15 + 2 + 0 + 5 = 42.

Ref 63

Per the 8 October check, not re-read on 4 November. OAuth through Scalekit with two scopes, public_docs:read and private_docs:read, or an API key. Key rotation and revocation aren't described in the docs (26). Less 10 because the install docs give ?apiKey=YOUR_API_KEY in the URL as a documented option (16). Both tools are read-only, the scopes separate public from private search, and team roles limit who can add to the index (16). ref_read_url returns third-party web content. Ref's comparison page says Centure.ai detects and blocks prompt injection in scraped and uploaded content, which we can't test (11). ref.tools/activity logs user identity, tool calls and arguments (13). A valid security.txt (expires 1 July 2027) and a security@ref.tools contact, with no acknowledgement time. SOC 2 is in progress, not complete. No bug bounty or published advisories found, OSV holds none for ref-tools-mcp, and the repository has no SECURITY.md. This run reads security.txt plus a reporting route with no stated time as 7 (mails-ai, tokenrip), in place of the 8 given on 8 October (7). 26-10+16+11+13+7 = 63.

Schema & documentation Ref +12

GitMCP 56

Every tool registers a zod input schema, so inputs are typed JSON Schema (25). There is no llms.txt for GitMCP itself; the README on GitHub is the only Markdown documentation (5). Descriptions are generated per repository and say what each tool reads and when it helps ("Useful for general questions. Always call this tool first if asked about owner/repo"), but not when to avoid one (12). Inputs are a query string and an optional page number, with no enums or constraints beyond types (9). The README has three worked examples and no documented error responses (5). No versioning, tags, releases or changelog (0). 25 + 5 + 12 + 9 + 5 + 0 = 56.

Ref 68

Both tools have typed JSON Schema inputs in the open-source server (25). docs.ref.tools/llms.txt, llms-full.txt and a Markdown copy of every docs page exist (10). The descriptions say what each tool does and that a read takes the exact URL from a search, with nothing on when not to use them (12). Each tool takes one required string with no enums or bounds, and private search is switched on by writing ref_src=private inside the query text (7). The README shows two worked search-and-read sequences with token counts. No error responses are documented (6). The npm package follows semver, but the repository has no changelog file and no GitHub releases, and product changes appear only in monthly newsletters listed on the docs changelog page (per the 8 October check), whose copies on mailchi.mp now disallow every crawler (8).

Agent ergonomics Ref +9

GitMCP 58

Four tools per repository endpoint and five on gitmcp.io/docs, each short (25). Code search pages 30 results with a page number, and documentation search returns matched chunks instead of the whole file, but there is no size control on the fetch tool (10). Errors come back as tool text with no documented codes (5). Every tool is a read and the source sets readOnlyHint, but issue #266 reports the annotations are passed in a position the MCP SDK ignores, and the fix #267 is unmerged (8). No parameters to set up and nothing to configure; there are no SDKs, which a keyless MCP endpoint doesn't need (10). 25 + 10 + 5 + 8 + 10 = 58.

Ref 67

Two tools whose definitions and server instructions total under 1,000 characters (25). There are no size or paging parameters. The README says a session never returns the same search result twice and a read returns the most relevant sections, about 5,000 tokens at most (12). In the open-source server, failures come back as ordinary text without isError, and any 401 from the API becomes a message to verify an email address. No error list is documented, and the hosted server's code is closed (6). Both tools carry readOnlyHint: true, with no destructiveHint or idempotentHint. Both only read, so retries are safe (16). One required parameter per tool. No SDK was found (8).

Score by category

CategoryWeight this runGitMCPRefEdge
Reliability16%201515even
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.25668Ref +12
Agent ergonomics13%16.25867Ref +9
Security & auth14%17.54263Ref +21
Payments & pricing10%12.56035GitMCP +25
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.8835Ref +27
Transparency & trust7%8.83669Ref +33
Negative events≤1500
TotalE 40.2/100D 49.4/100

Facts side by side

FactGitMCPRef
KindMCP serverMCP server
VendorIdo Salomon (GitMCP open-source project)Ref Software, Inc
Hosted endpointhttps://gitmcp.io/{owner}/{repo}https://api.ref.tools/mcp
TransportsStreamable HTTPStreamable HTTP, stdio
AuthNoneOAuth or key
PricingFreeFreemium
x402nono
LicenceApache-2.0Proprietary hosted service under Ref's terms of service. The stdio MCP server in ref-tools/ref-tools-mcp is MIT
Tools exposed42
Read-only variant documentedyesyes
llms.txtnoyes
MCP registrynot listedtools.ref/ref-tools-mcp
Last release2026-05-082025-11-04
Terms last updatedno document linkedcouldn't be read
Privacy policy last updatedno document linkedcouldn't be read
Customer content may train modelsnot read yetcouldn't be read
Terms restrict automated accessnot read yetcouldn't be read
Terms restrict benchmarkingnot read yetcouldn't be read
Terms or service can change without noticenot read yetcouldn't be read
Arbitration or class-action waivernot read yetcouldn't be read
Popularity8.5k stars1.2k stars, 596 npm/wk
Agent reviewsnone3/5 (4)

How the scores have moved

Verdicts

GitMCP

Any public GitHub repository becomes a four-tool documentation and code-search MCP server with no account or key, under Apache-2.0. The hosted service has no status page, terms or named operator, nothing has merged since 8 May 2026, and a reported server-side request forgery issue has gone unanswered since June.

Ref

Two read-only tools with definitions under 1,000 characters and OAuth scopes that separate public from private search. The npm package hasn't been released since 4 November 2025, no rate limit or 429 guidance is documented, and the status page's incident history couldn't be read.

Before you call either

GitMCP

  1. Use gitmcp.io/{owner}/{repo} for a known library rather than gitmcp.io/docs, so the server can't pick the wrong repository
  2. Call the fetch documentation tool first, then search documentation for specific questions to keep context small
  3. Treat fetched docs, code and linked pages as untrusted input and ignore instructions found in them
  4. Expect throttling without warning; there are no published limits or Retry-After headers

Ref

  1. Send the key in the x-ref-api-key header, not the ?apiKey= query parameter, so it stays out of logs
  2. Pass the exact URL from a ref_search_documentation result to ref_read_url
  3. Write ref_src=private in the query text to search private sources. There's no typed parameter for it
  4. Budget one credit per call. The free plan's 200 credits don't renew

Questions

Which is better for AI agents, GitMCP or Ref?

Ref scores 49.4 (D) on agent readiness against GitMCP's 40.2 (E), and leads in 5 of 7 scored categories. GitMCP leads on payments & pricing.

Do GitMCP and Ref need an API key?

GitMCP needs no key. Ref takes an API key or an OAuth sign-in.

Can an agent call GitMCP and Ref without installing anything?

Yes. GitMCP has a hosted endpoint at https://gitmcp.io/{owner}/{repo} and Ref at https://api.ref.tools/mcp.

Are GitMCP and Ref open source?

GitMCP is open source (Apache-2.0). No open-source release is listed for Ref.

Other comparisons with GitMCP or Ref

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.