One platform. Every review surface.

A well-tuned Mesrai review catches 70-90% of real issues with a false-positive rate under 10% on most codebases. Accuracy improves over time as Mesrai Rules tune themselves from your team's reactions to past findings — bot accuracy is a moving target, but the trend is consistently upward inside any given workspace.

Mesrai Rules are team-specific review rules written in plain English or YAML — things like "always use the structured logger, never console.log" or "tests must live next to their source file". Mesrai applies them on every PR alongside the built-in bug, security, and performance agents.

Yes — BYOK is supported on every paid plan. Connect OpenAI, Anthropic, Google Vertex, AWS Bedrock, Groq, Cerebras, or any OpenAI-compatible endpoint. Mesrai charges a flat platform fee; your LLM bill flows directly to your provider at their rates — Mesrai never marks up tokens.

Yes. The mesrai-graph AST engine builds a semantic index across the whole repository, so the review understands cross-package dependencies, type imports, and architectural boundaries — not just the diff in isolation. Tested on monorepos of 1M+ LOC.

Three ways. (1) The cloud reviewer runs automatically on every PR via the git-provider webhook. (2) The Mesrai CLI (`npx mesrai review`) can run as a CI step inside GitHub Actions, GitLab CI, CircleCI, and similar. (3) The pre-push hook gives local-machine review before the PR even opens.

Yes. 14 days, every Pro feature unlocked — bring your own AI key. No credit card required, no auto-charge at the end. Sign up at app.mesrai.com.