Last updated: 2026-08-12
Greptile is an AI code review agent used by more than 22,000 engineering teams, including PostHog, Retool, and Nvidia. It builds a semantic graph index of an entire codebase before reviewing a pull request, tracing dependencies and git history across files to flag cross-module bugs that diff-only reviewers miss.
About Greptile
Greptile is an AI code review agent built by the San Francisco startup of the same name, founded in 2023 by Soohoon Choi, Vaishant Kameswaran, and Daksh Gupta and backed by Y Combinator's Winter 2024 batch. Rather than reviewing only the lines a pull request touches, it constructs a graph-based map of the whole repository, indexing functions, classes, import chains, and commit history before posting a single comment. Customers include PostHog, Retool, Klaviyo, Nvidia, Brex, Scale, and Zapier. Its v4 review agent, released in March 2026, replaced a single-pass diff scan with a multi-hop reasoning loop that traces dependencies and git history across files before flagging an issue, and each comment carries a confidence score so developers can triage quickly. Greptile's own changelog reports the change lifted addressed-comments-per-PR 74%, and author reply and upvote rates rose alongside it, evidence the deeper context produces feedback developers actually act on. Teams scope what gets reviewed through a .greptile.yaml configuration file, a linked markdown style guide, or rules written for specific paths in plain English, and Greptile adapts later reviews as it observes how a team responds, reacts, and commits over time. It installs as a GitHub App or GitLab integration and posts inline PR comments automatically, and a companion MCP server hands review data and custom context to editor-based agents so they can act on flagged issues in place. It fits scaling engineering teams on GitHub or GitLab willing to tune rules to manage noise in exchange for catching cross-module bugs; teams on Bitbucket or Azure DevOps, or those wanting fast single-pass review, are better served elsewhere. Pricing follows a free-plus-credits model: a no-cost tier for a single developer, a per-seat paid plan for teams, and custom enterprise terms that add self-hosted deployment inside a customer's own AWS environment with a bring-your-own-LLM option, SSO/SAML, and GitHub Enterprise Server support. Greptile is web-based, run through app.greptile.com and the GitHub/GitLab integrations, with no native desktop or mobile app. Greptile raised a $4.1M seed round in June 2024 led by Initialized Capital, then a $25M Series A in September 2025 led by Benchmark Capital, reportedly valuing the company near $180M. It is SOC 2 Type II compliant, with reports available to customers on request, and customer data is hard-deleted from production systems within a day of an admin-initiated request.
Pricing
Starter is free for one active developer with unlimited repositories and 50 credits a month (1 credit per standard review, 3 per TREX review); Pro costs $30 per seat per month with 50 credits per seat, unlimited seats, and $1 per extra credit; Enterprise is custom-priced and adds self-hosted deployment. Open-source projects under MIT/Apache are free, and pre-Series-A startups under $2M revenue get 50% off.
Key Features
- Semantic Graph Index: Builds a graph index of every function, class, import, and commit in a repository before reviewing a pull request, so comments account for cross-file impact instead of just the diff.
- v4 Multi-Hop Agent: The March 2026 v4 agent traces dependencies and git history across files in a reasoning loop, lifting addressed-comments-per-PR by 74% over the prior single-pass version.
- Confidence-Scored Comments: Every review comment carries a confidence score so developers know which flags to act on first.
- Custom Rules in Plain English or YAML: Teams define review directives through a .greptile.yaml file, a linked markdown style guide, or scoped plain-English rules for specific paths.
- Learns From Team Feedback: A memory layer watches how engineers respond to past comments, reactions, and commits, and tunes what it flags next time.
- Open MCP Server: An MCP server hands Claude Code, Cursor, VS Code, and Codex CLI direct access to PR data, review comments, and custom context so those agents can act on flagged issues in the editor.
- Self-Hosted Enterprise Deployment: Enterprise customers can run Greptile inside their own AWS environment with their own LLM provider for air-gapped review.
Pros
- One independent benchmark found an 82% bug catch rate, nearly double CodeRabbit's 44% in the same test.
- Traces dependencies and git history across an entire repository, catching cross-module bugs and dependency breaks that diff-only reviewers miss.
- Free Starter tier includes unlimited repositories, so a solo developer can try full-codebase review with no card on file.
- Ships an open MCP server, so tools like Claude Code and Cursor can query review comments and apply fixes without a context switch.
Cons
- Full-context reviews take several minutes per pull request, slower than diff-only tools with near-instant turnaround.
- The same independent benchmark recorded 11 false positives per review run, more noise than lighter competitors, so rules need tuning.
- Repository support is limited to GitHub and GitLab; there is no Bitbucket or Azure DevOps integration.
- No public GDPR, HIPAA, or ISO 27001 attestation is published; only SOC 2 Type II compliance is confirmed on Greptile's security page.
Frequently Asked Questions
How much does Greptile cost in 2026?
Pro tier billing runs $30 per seat each month and bundles 50 review credits into that price, with extra credits billed at $1 apiece. Anyone starting out gets the Starter tier at zero cost, covering a single developer and 50 monthly credits across unlimited repos. Larger organizations move to Enterprise, a custom quote that layers in self-hosting, SSO/SAML, and a dedicated Slack channel. Non-profit open-source work under MIT or Apache terms rides free, and early-stage companies below $2M in revenue qualify for half off.
Is Greptile free to use?
Yes, Greptile ships a no-cost Starter tier scoped to a single developer, unlimited repos, and 50 credits every month, with no payment card needed to begin. Growing past one seat or burning through the monthly credit allowance means moving to the $30-per-seat Pro tier. Custom review rules and unrestricted external app connections stay behind that Pro paywall and are absent from Starter.
What are the best alternatives to Greptile?
CodeRabbit is the most widely deployed AI reviewer and, in one independent benchmark, produced far fewer false positives than Greptile, making it a better fit for teams that prioritize low noise over raw catch rate. Qodo Merge runs a multi-agent review architecture across GitHub, GitLab, Bitbucket, and Azure DevOps, useful for teams on platforms Greptile does not support. Cursor's BugBot is a flat-fee option built to flag only clear-cut bugs with minimal commentary, for teams that find AI review comments too noisy.
How does Greptile compare to CodeRabbit in 2026?
One independent benchmark (50 real pull requests across open-source projects including Sentry, Cal.com, and Grafana) found Greptile catching 82% of injected bugs versus CodeRabbit's 44%, because Greptile traces dependencies and git history across the whole repository rather than scanning only the diff. The same benchmark found Greptile producing 11 false positives to CodeRabbit's 2, and full-context reviews run several minutes instead of CodeRabbit's faster turnaround. Choose Greptile for complex, multi-service codebases where a missed bug is costly; choose CodeRabbit when review speed and low comment noise matter more.
How do you get started with Greptile?
Sign up at app.greptile.com and install the Greptile GitHub App, or connect a GitLab access token to a repository; the free Starter tier activates immediately with no card required. From there, add a .greptile.yaml file or plain-English review rules to scope what Greptile checks, and connect its MCP server to Claude Code, Cursor, VS Code, or Codex CLI to act on review comments inside the editor.
Top Alternatives
- Qodo: Pick Qodo Merge if you need review coverage on Bitbucket or Azure DevOps too; pick Greptile if you're on GitHub or GitLab and want deeper cross-file bug tracing.
- Cursor: Pick Cursor's BugBot if pull request comment noise is your main complaint; pick Greptile if you'd rather catch more real bugs and tolerate more false positives.
- Sourcegraph Cody: Pick Cody if you want an interactive assistant for exploring a large multi-repo codebase; pick Greptile if you want automated bug-catching on every pull request.