Help your agentscoworkersgrandmateenagerteamdogselfcathamsterwifehusband write high quality Rust code.
AI agents can write faster than anyone can review. Now you can keep up. cqx can process 100,000 lines in less than a second. It gives you real insight into what your agents are building & actionable intelligence when slop happens. It even runs in the browser and can analyze public GitHub repos in a matter of seconds. Try it out yourself below.
crates/host/src/modules/fs.rs:81 let _ = enforce_read(Some(&base)); ^^^^^^^^^^^^ called, and its answer thrown away
crates/runtime/src/command_run.rs:294 Command::new("cmd").args(["/C", command]) ^^^^^^^ a shell, handed an argument nothing resolved
crates/harness/src/command.rs:74 .arg("--allow-all") ^^^^^^^^^^^^^ the child spawned with its own checks off
| containment | 70 |
| legibility | 74 |
| modularity | 100 |
| quality | 28 |
| security | 58 |
Nothing to install. Nothing to build.
Compiled to WebAssembly and run in your browser, across four threads. Your code is never uploaded, and a repository that does not compile still reads.
bevy, read cold. The second visit pays none of the fetch — a file is named by its contents, so what has not changed is already here.
- No clone
- Files come from the CDN.
- No build
- It reads source, not artefacts.
- 1.9M lines
- makepad, analysed in 12.8s.
A compiler's answer, not a checklist.
Each finding names a file, a line and the code on it, with the reason it matters and what to do instead. Five categories, each configurable per repository, so the standard is the one you set.
It does not replace clippy, cargo audit or cargo deny, and it does not try to. Those read a file, a lockfile and an advisory feed — and they do it better than we would. cqx reads what is between them: what calls what, which paths reach a check, and what your program can actually do once it runs. Keep all four.