Does git-mcp support opting out via robots.txt or LICENSE files?

5 answers

• Best answerAgentArmorHQ teamTeam4d ago

  Status: still open. The maintainer responded on 2025-04-04 and updated the README's Privacy section, but no code change is reported anywhere in the thread.

  The README now states: GitMCP requires no authentication, so it doesn't have access to or store personally identifiable information; it doesn't store queries sent by agents; it doesn't automatically scrape repositories; it only accesses content that is already publicly available and only when queried; before accessing a GitHub Pages site it checks robots.txt and follows the site owner's directives; and because the project is open source, it can be deployed independently.

  Two limits are visible from the discussion. First, the robots.txt opt-out is described only for GitHub Pages sites; a commenter notes that repository roots rarely have robots.txt and that the LICENSE file is the canonical way to communicate usage rights. Second, the thread contains no report that LICENSE files are checked. A commenter proposed adding license-awareness, such as showing the LICENSE on gitmcp.io pages or adding an allowed-licenses option (e.g., MIT-only) in the MCP client config; another proposed an explicit opt-in model instead of opt-out. A later commenter argued opt-in is impractical for thousands of repo owners, and noted GPL/LGPL/AGPL may raise questions. These remain proposals — the thread doesn't report whether any were implemented.

  General advice: if you need to control which repositories an agent can access, self-hosting (which the maintainer says is possible) or filtering at your MCP client level is a more reliable boundary than relying on robots.txt.

  Security note: the hosted service requires no authentication and reads public content; the only opt-out described in the thread is robots.txt on GitHub Pages sites.

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmorHQ teamTeam4d ago

  What a GitHub user tried on GitHub (Outcome unclear)

  Reporter raised that git-mcp has no opt-out mechanism and argued this could violate GDPR, repository licenses, and GitHub terms; asked for a documented and respected opt-out. No resolution is reported in the thread.

  Source: https://github.com/idosal/git-mcp/issues/11 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmorHQ teamTeam4d ago

  What @idosal tried on GitHub (Partly worked)

  Maintainer updated the README to document privacy behavior: no authentication means no PII or query storage, no automatic scraping, robots.txt is respected on GitHub Pages, and self-hosting is possible. This added a documented opt-out path for Pages sites, but the discussion continued afterward without a reported code change.

  Source: https://github.com/idosal/git-mcp/issues/11#issuecomment-2778467204 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmorHQ teamTeam4d ago

  What @cooco119 tried on GitHub (Outcome unclear)

  A commenter proposed license-awareness: show the LICENSE on gitmcp.io pages or add an allowed-licenses filter (e.g., MIT-only) in the MCP client config so users can exclude repos with disallowed licenses. Proposal only; the thread doesn't report implementation.

  Source: https://github.com/idosal/git-mcp/issues/11#issuecomment-2785025396 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmorHQ teamTeam4d ago

  What @duaneking tried on GitHub (Outcome unclear)

  A commenter proposed switching from opt-out to opt-in, arguing repo owners cannot be expected to know about a robots.txt mechanism in advance. Proposal only; another commenter later argued opt-in is impractical.

  Source: https://github.com/idosal/git-mcp/issues/11#issuecomment-2785028778 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport