Reviews from your terminal.

• mesrai auth loginAuthorize the CLI to your workspace
• mesrai reviewReview the current diff against base branch
• mesrai review --files src/api/users.tsReview specific files only
• mesrai review --severity highSet minimum severity threshold for output
• mesrai rules syncPull latest Mesrai Rules from your workspace

mesrai skills install auto-detects every agent root present on the machine and writes its bundled skills into each one. A `.mesrai-managed-skills.json` manifest tracks which files this run owns — re-syncs are idempotent, uninstalls leave nothing behind.

• Claude Code
• Cursor
• Codex
• Windsurf
• OpenCode
• Goose
• Gemini CLI
• Kilo Code
• Roo Code
• AiderDesk
• Agents
• Antigravity
• Droid
• Kiro
• + more local roots

1. 01

   Single installer

   One install URL updates the CLI and writes bundled Mesrai skills into every detected agent root on the machine — no separate repos, no per-IDE setup.

2. 02

   Agent-friendly output

   `--agent`, `mesrai schema`, and `--fields` give skills, CI runners, and thin MCP layers a stable JSON envelope they can parse without scraping terminal output.

3. 03

   Local + remote review flows

   `mesrai review` for local diffs against your team's Mesrai Rules. `mesrai pr suggestions` for fetching server-side suggestions on an existing GitHub/GitLab PR.

• How do I install the Mesrai CLI?+

  Two ways. (1) `npx mesrai` runs it without installing anything — best for one-off reviews or CI steps. (2) `npm install -g @mesrai/cli` installs it globally and exposes the `mesrai` binary. Both ship the same engine; pick based on whether you want the binary persistent on your machine.

• Can I run reviews on my local machine before opening a PR?+

  Yes — that's the primary use case. `mesrai review` from the repo root reviews uncommitted + unpushed changes. Pair it with a Git pre-push hook (`mesrai install hook`) and every push is reviewed before it leaves your machine. Same engine as the cloud reviewer; same Mesrai Rules applied.

• Does the CLI work in CI (GitHub Actions, GitLab CI, CircleCI)?+

  Yes. Add `npx mesrai review --ci` as a step in any CI runner — it auto-detects the PR base + head from the standard CI environment variables (GitHub Actions, GitLab CI, CircleCI, Buildkite, Jenkins). The CLI exits with a non-zero code if blocking-severity findings exist, so it slots cleanly into a required-check gate.

• Where does the CLI store its config and credentials?+

  Config lives at `~/.config/mesrai/config.json` (Linux/macOS) or `%APPDATA%\mesrai\config.json` (Windows). Auth tokens are stored in the OS keychain (macOS Keychain, GNOME Keyring, Windows Credential Manager) — never in plaintext on disk. Rotate or revoke from Settings → CLI Tokens in the web app.

• Can I review a specific branch or commit range?+

  Yes. `mesrai review --base main --head feature/x` reviews the diff between any two refs. `mesrai review --since HEAD~10` reviews the last 10 commits. `mesrai review path/to/file.ts` reviews a single file. Useful for batch-reviewing legacy branches or auditing a specific change set.

• Does the CLI work offline or without internet?+

  Partially. The CLI ships an offline ruleset (~120 deterministic rules + lint checks) that runs entirely locally — no network required. The full multi-agent AI review needs a network call to either the Mesrai cloud (default) or your own LLM endpoint (BYOK). Local-only mode is enabled with `mesrai review --offline`.