Greptile
An AI code review agent that reviews pull requests with full-codebase context — catching multi-file logical bugs and learning your team's standards.
Greptile reviews pull requests with full context of the codebase — not just the diff — so it catches multi-file logical bugs diff-scoped reviewers miss. It learns team standards from your engineers' own PR comments, takes custom rules in plain English (and reads CLAUDE.md/.cursorrules), and hands fixes off to Claude Code or Cursor. Free Starter tier; Pro per seat, 14-day trial; free for OSS.
Greptile attacks the weakness of diff-scoped review: most real bugs aren't visible in the changed lines alone. It indexes your entire codebase and reviews every pull request against that context — the callers of the function you changed, the convention the rest of the module follows, the invariant two files away — which is how it catches the multi-file logical bugs that pattern-matching reviewers wave through.
Highlights
- Full-codebase context — reviews reason over the repository, not the diff, targeting cross-file logic errors.
- Learns your standards — the v4 agent architecture (March 2026) learns from your engineers' own PR comments, so the bot's taste converges on the team's and nitpick noise drops.
- Rules in plain English — codify standards conversationally; it also auto-detects existing
CLAUDE.mdand.cursorrulesfiles as conventions. - Agent handoff — "Fix with your Agent" sends findings straight to Claude Code, Cursor, or Codex; an MCP server exposes reviews and patterns inside those tools.
- CLI for local review —
npm i -g greptilereviews branches from the terminal before a PR exists. - Enterprise posture — SOC 2 Type II, SSO, audit logs, and self-hosting (Docker/Kubernetes, air-gapped with custom LLMs).
In an AI-assisted workflow
Install the GitHub or GitLab app, select repos, and reviews start appearing on PRs within minutes. The compounding loop is the point: agents write more of the code, Greptile reviews it with repo-wide context, and its findings route back to the agent that wrote it — closing the write → review → fix cycle without a human ferrying comments.
TIP
Treat the learning period deliberately: keep human review on high-stakes paths for the first weeks while Greptile absorbs your team's comment patterns — its precision visibly improves as it learns what your engineers actually flag.
Good to know
Greptile is a proprietary SaaS (the GitHub org hosts integrations, not the product), backed by a $25M Series A led by Benchmark (September 2025) and used by 9,000+ teams including Brex and PostHog by mid-2026. Reviews are metered in credits per seat (50/month on Pro, then $1 per extra credit as of September 2026, with a TREX runtime-validation review costing 3) — relevant for very high-PR-volume teams; a free Starter tier covers a single active developer. GitHub and GitLab only; Bitbucket/Azure DevOps shops should look at Qodo. How it stacks against the field is in Best AI Code Review Tools in 2026.
Frequently asked questions
- What makes Greptile different from other AI code reviewers?
- Codebase-wide context and learned standards. It indexes the whole repository, so a review considers callers, conventions, and side effects beyond the diff — and its v4 architecture learns from how your engineers actually comment on PRs, cutting nitpick noise over time. Custom rules are plain English, and it auto-detects CLAUDE.md and .cursorrules files.
- How much does Greptile cost?
- Plans, as of September 2026 from greptile.com/pricing: Starter is free for one active developer with unlimited repositories and 50 credits a month; Pro is $30 per seat/month with 50 credits included per seat and $1 per additional credit; Enterprise is custom-priced. Credits meter reviews — one standard review costs 1 credit and one TREX review costs 3. Pro has a 14-day free trial, pre-Series A startups under $2M revenue get 50% off, and qualifying non-commercial open-source projects (MIT or Apache) are free.
- Does Greptile work with GitLab and self-hosting?
- Yes on both: GitHub (Cloud and Enterprise Server) and GitLab (Cloud and Self-Managed) are supported, and the platform is SOC 2 Type II with self-hosted deployment via Docker Compose or Kubernetes — including air-gapped setups with custom LLMs for stricter environments.
Filed under
code-review · ai-review · pull-requests · ci
Related
- QodoA quality-first AI code review platform (ex-CodiumAI) — multi-agent PR review with your team's rules, plus IDE, CLI, and codebase-intelligence products.
- Best AI Code Review Tools in 2026The AI code reviewers worth running in 2026 — CodeRabbit, Greptile, Qodo, Cursor Bugbot, and GitHub Copilot code review compared, with a verdict for each.
- CoderabbitAn AI code reviewer that posts line-by-line feedback and summaries on every pull request.
- Code ReviewerUse this agent to review code changes for correctness, security, and maintainability before merging. Examples — reviewing a PR diff, auditing a new module, checking a refactor for regressions.
- Review PRReview a pull request for correctness, security, and style, and summarize findings.
- Running Claude Code in CI: Headless Mode & GitHub ActionsClaude Code without the terminal — claude -p flags, JSON and structured output, safe permission scoping, and the official GitHub Action responding to @claude.
- An AI Code Review Workflow That Actually Catches BugsLayer the review stack — self-review, AI reviewers, tests, and a human pass focused on what machines miss — into a workflow tuned for AI-written code.