Analyze a GitHub repository
Paste the URL of a public GitHub repository. RepoBridge fetches its file tree, decides which files are worth reading and why, reads and normalises those it can, writes a summary grounded only in what it actually read, and records what happened at every stage — including what it skipped and what failed.
Try https://github.com/psf/requests if you want something to paste. A repository this size takes roughly ten seconds.
Configuration
This instance runs on the operator’s API keys. Nothing here is yours to set, and no key is ever entered in the browser.
-
configured
REPOBRIDGE_GITHUB_TOKEN
GitHub requests are authenticated: 5,000 per hour. One analysis costs two requests plus one per file read.
-
configured
REPOBRIDGE_GEMINI_API_KEY
Summaries are generated by gemini-3.5-flash-lite. The free-tier daily quota is per model; set REPOBRIDGE_GEMINI_MODEL to switch.
Read from the environment only, never from a form field or a request, and never displayed here — this panel reports only whether each one is set.
Limits on this instance
- 20 analyses per hour, everyone combined
- 5 analyses per hour per visitor
- repositories up to 5,000 tree entries
- at most 100 files read per analysis — larger repositories are sampled, and every run says how many of how many it read
The limits exist because every analysis spends the operator’s GitHub and summarisation quota. When one is reached the page says which, and when it lifts.
Recent runs
Showing 2 of 2 stored runs.
Stored runs do not survive a restart on this instance.
| Repository | State | Entries | Selected | Read | When | Views |
|---|---|---|---|---|---|---|
| karpathy/nanoGPT | completed | 32 | 24 | 24 | 2026-08-26 10:40 UTC | files for karpathy/nanoGPT diagnostics for karpathy/nanoGPT |
| mungell/awesome-for-beginners | completed | 13 | 9 | 9 | 2026-08-26 09:23 UTC | files for mungell/awesome-for-beginners diagnostics for mungell/awesome-for-beginners |