Repository navigation
Five specialized LLM agents review each PR and post one prioritized, transparent comment. Runs in your CI, BYOK (OpenAI / DeepSeek / OpenRouter / Groq / Ollama). Injection-hardened, never breaks your CI. Eval: 34/35 fixture-runs, zero clean-fixture false positives.