The three tools split by scope — everything, one directory, one file — which is what stops a large repository from arriving in a single response. There is no search and no history here: you navigate by path, so it works best when you already know roughly where in the tree to look.
A three-tool reader over a GitHub repository addressed by repoUrl: the whole directory listing, one directory, or one file.
- Every directory in a repository, from its URL alone
- The entries under one path, for when the full tree is more than the question needs
- The contents of a single file by path
No account and no key. Each call carries the repoUrl, so the repository has to be one that can be read without authentication, and you supply the URL rather than configuring a default.
One command — npx github-repo-mcp
