A watertight compartment for your codebase. Isolate bad code, measure it against rules, ship a priority-sorted verdict. Code management for the AI age.
Cofferdam documents customs and rules that you want to apply across the codebase. Agents can write code and at compile time ensure it compiles with your wider codebase without having to read the entire context.
Cofferdam is how you're able to apply standards on security, architecture and more to your codebase. Apps can be generated rapidly and Cofferdam is how you apply your standards.
Brings the five-category discipline that Credo made popular — Warning, Refactor, Design, Readability, Consistency — to TypeScript. If you've used Credo, the category names and report shape will feel familiar.
Priority and severity are separate axes. Priority is computed; severity is configured. The report sorts by what to fix first, and CI gates on what must not regress.
Adopt on a legacy codebase without drowning in noise — default mode shows only new findings. Capture the current state with `cofferdam init --baseline`, then tighten over time.
SARIF output, `--since main` for PR-only mode, GitHub annotations, and a `--robot` flag for machine-readable compact output that AI agents can consume directly.
`cofferdam advise` emits the rules that apply to a file before any edit, so an LLM agent can plan against your layering and complexity limits instead of reverse-engineering them from violations. Agent advisory →