You know that thing where you clone a repo and spend the first 15–20 minutes just trying to figure out what it even is?
You open package.json, look for scripts, check if there's a Dockerfile, search for the entry point, wonder if there are tests anywhere. And you do this every single time.
I got annoyed enough to build something about it. It's called codeglance.
What it does
Run one command in any repo:
npx codeglance
You get a quick summary: what the project is, how to run it, which files to read first, and what tooling is set up. No config, no API keys, no setup.
It reads your manifest files: package.json, go.mod, Cargo.toml, pyproject.toml : and figures out the stack from there. Not just "this is a TypeScript project" but "this is Next.js 14 with Prisma and tRPC."
The "where to start" section ranks files by depth, naming patterns, and known entry point conventions. Heuristic, not magic: but usually right enough to be useful.
There's also a --for-ai flag that generates a compact summary you can paste into Claude or Cursor before asking questions about a codebase. Saves you from dumping the whole repo in.
What it doesn't do
It doesn't parse source code or understand logic. If a project has no standard manifest, the output will be shallow. File ranking is pattern-based, not import-graph analysis — so it's approximate. Java and Ruby aren't supported yet (on the roadmap).
How it's built
TypeScript, Commander for CLI parsing, tsup for bundling. 70 tests. Supports Node.js, Python, Go, Rust, and C/C++ repos. I used Claude Code heavily during development and I'm not pretending otherwise: I care more about whether the output is useful than getting credit for typing every line.
Try it
npx codeglance
npx codeglance ~/projects/some-repo
Repo: https://github.com/mansoor-mamnoon/codeglance
If the "files to read first" section is useless on your repos, I genuinely want to know: that's the part I'm least confident about. ⭐ if it saves you time.