Your team's engineering standards, surfaced as you type, checked before review, and shipped in every PR.
Every team has a way of doing things — and most of it lives in people's heads, in review comments, in a CONTRIBUTING.md nobody reads. Quack turned those expectations into structured guidelines and served them back inside the editor, exactly where the code gets written.
Apache 2.0 · self-hostable · runs on local LLMs
A loop between the people who set the bar and the people who ship
Guidelines were never a document. They were a living index that the IDE could query, the model could reason over, and maintainers could measure.
Curation
Maintainers wrote their expectations once — naming, error handling, testing, review etiquette. Quack parsed existing docs and past PR comments to bootstrap the list.
Assistance
The extension pulled only the guidelines relevant to the file you had open, checked your diff against them, and generated compliant code — before a reviewer ever saw it.
Iteration
Which rules got broken most? Which ones nobody understood? Maintainers saw where the contribution flow leaked, and rewrote the guidelines that weren't landing.
Built on open models, not someone else's cloud
Quack ran against Ollama and any OSS model you pointed it at. Self-host the whole stack, keep every line of source on your own infrastructure.
The product is sunsetted. The code isn't.
Everything we built is Apache 2.0 on GitHub — fork it, run it, take the parts you need.
Browse the organisation →New problem, same team
Quack AI is no longer maintained. After a stint on SourcePulse, the founders left developer tooling behind and now build Rimward, security monitoring for solar sites.
Visit rimward.ai →
