Code Terrier — Every PR reviewed by many models in parallel

Every pull request read by more than one model — each checking the others' work.

Agents now write more code than anyone can read. Code Terrier reads it. Each pull request is reviewed inside a real coding agent that opens files and follows the change, not a one-shot pass over the diff.

• Cross-model verification. Run up to three reviewer models side by side — Claude, GPT, Gemini, DeepSeek, Kimi, Qwen, GLM, Grok — and every finding is graded by a model that did not raise it. One model's blind spot is another's catch, and disagreement drops false positives before they reach your PR.
• Inline and categorized. Security, Bug and Flag findings anchored to the exact line, not a wall of prose at the bottom of the thread.
• Fixes you can commit. Findings arrive as GitHub suggestions, plus a non-blocking commit status on every head commit.
• Reply and it re-checks. Answer a review comment and the loop runs again: what you fixed is marked resolved, what you pushed is re-read.

Your keys, or none. Point it at your own Anthropic, OpenAI, Google, or any OpenAI-compatible endpoint — or run on the shared pool with no key at all.
Install once and every PR is reviewed automatically. Prefer to ask rather than be told? Switch to first-look-only or fully manual, and comment @code-terrier review when you want one.