Skip to main content

AI Code Review

Greptile

AI code review agent that reviews every pull request against a graph of the whole repository, in Greptile's cloud or self-hosted.

The 30-Second Take

Pick Greptile if review queues outgrow your reviewers and bugs cross file and service boundaries.

It reviews every pull request against a graph of the whole repository and posts comments with suggested fixes. Most teams that wrote publicly about it kept it.

Best fit
Pull request review on busy repositories where bugs cross file and service boundaries.
Main constraint
Spend grows with every review. Deeper tiers cost more credits, and every author and agent draws on the same prepaid pool.
First move
Run Pro on one busy repository with automatic recharge off and bot authors excluded.

Why

Watch the pool: every review on Pro draws on one prepaid organization pool at US$1 a credit, and deeper tiers cost more. Tune strictness, comment filters and custom rules over the first weeks to keep noise down.

Practitioner sentiment, leaders and teams+40 Positive

20 practitioners, 66 sources reviewed, Updated Oct 25, 2026How this is measured

Read the GILD Review
Save a product your organization uses.Keep up with Greptile without adding it to your stack.Visit Website
Operating stack
Agentic Software Development

Comparisons

How does Greptile compare with the alternatives?

See all 4 comparisons

  1. Greptile vs CodeRabbit

    • AI Code Review

    Choose Greptile for fewer noisy comments and review depth set per pull request, and CodeRabbit for Azure DevOps or a review extension in the IDE. Billing differs too: prepaid credits shared across the organization against seats with hourly limits.

  2. Greptile vs GitHub Copilot

    • AI Code Review

    Start with Copilot code review if your code is on GitHub, since your Copilot plan includes it. Choose Greptile when code also sits on GitLab, Bitbucket, Gitea, or self-managed hosts, when review must be self-hosted, or when you want fewer noisy comments.

  3. Greptile vs Qodo

    • AI Code Review

    Both bill reviews from one shared credit pool with no seat fees. Choose Greptile for prepaid US$1 credits with no plan fee, and Qodo for a monthly credit pack. Check Git hosts first: Greptile adds Gitea and Perforce, and Qodo adds Azure DevOps and Gerrit.

Buying and deployment

What engineering leaders should verify before a broader rollout.

Sourced from Greptile documentation

How you buy it

US$1 a credit

Pro has no seat fees. Credits are shared across the organization, with a 14-day trial.

Source 1, Greptile, Pricing: PricingSource 2, Greptile, Billing: Billing

Details

Starter is free for one active developer with 50 credits a month. Pro has no seat fees: credits cost US$1 each, shared across the organization by unlimited developers and agents, with a 14-day trial. Enterprise is custom.

Cost drivers

1, 3, or 10 credits

Base, Plus, and Apex reviews. Credits are prepaid, and new reviews pause at zero without recharge.

Source 2, Greptile, Billing: Billing

Details

A Base review uses 1 credit, Plus 3, and Apex 10, and a TREX test run adds 3. Only completed reviews spend credits. Credits are prepaid, and new reviews pause when the balance runs out unless automatic recharge is on.

Where it runs

GitHub, GitLab, Bitbucket

On Greptile Cloud. Enterprise adds self-hosting, air-gapped included.

Source 3, Greptile, Code providers: Code providersSource 5, Greptile, Deployment options: Deployment options

Details

Greptile Cloud supports GitHub, GitLab, Bitbucket Cloud, and Cursor Origin in beta. Enterprise adds self-managed hosts and Gitea, plus self-hosting on Docker Compose or Kubernetes, including air-gapped deployments.

What to verify

Training opt-out

De-identified customer data can train models until you opt out, and auto-approve needs no human.

Source 4, Greptile, Security practices: Security practicesSource 6, Greptile, Auto-approve PRs: Auto-approve PRs

Details

Help us improve Greptile is on by default, and the security page allows training on de-identified customer data until you opt out. Greptile Cloud keeps code cached until access is revoked, and auto-approve can approve pull requests without a human.

Signals

What is the current evidence saying about Greptile?

What engineering leaders, teams, and individual builders report about Greptile. Every finding links to its public sources.

Whose sources count

Practitioner sentiment, leaders and teams

+40 Positive net

of 20 sources that take a side, each author counted once. Small sample.

How this figure is made

Current sentiment

Positive, mixed and negative sources

Team and leadership accounts mostly keep Greptile. Eleven kept it, including Brex, which aligned on Greptile as its one agentic reviewer, and PostHog, which kept it while retiring CodeRabbit. Five ended a trial or moved to Codex, Cursor's Bugbot, Claude, or CodeRabbit. Every team account about bug detection reports real catches, including two serious bugs in an Appwrite pull request. Teams use Greptile before human approval. On comment noise, two team accounts raise it, and two, Brex's CTO among them, find the noise acceptable.

Run a pilot on live pull requests to measure which comments engineers act on and which they dismiss.

Evidence field. All 66 sources, one mark each.

Arrange by

Each mark is one source: positive, mixed, negative, no opinion stated, or documentation (not counted).

By sentiment, Engineering leaders and teams. Every source GILD read, grouped as the field draws it.
GroupSourcesPositiveMixedNegativeNo opinion stated and not counted in the scoreDocumentation, checked and not counted
Positive11110000
Mixed606000
Negative300300
Documentation18000018
Outside this group28147700
Every source GILD read, by the quarter it was published, as the field's time axis counts it. Documentation is dated by the day it was checked.
QuarterSources
Q1 20251
Q2 20252
Q3 20253
Q4 20253
Q1 202611
Q2 20265
Q3 202615
Q4 20268
Documentation, checked18

By sentiment

18 documentation pages check facts and are not counted; 28 sit outside this group.

Hover a mark for a quick look, click it or press Enter to open its source.Tap a mark to open its source, then step to its neighbors with Previous and Next.

What practitioners praise

Each mark is a source on it; filled marks praise it.

  1. Bug DetectionAll 8 sources on it praise it

Most common concerns

Each mark is a source on it; filled marks raise it.

For engineering leaders and teams, no concern reaches 5 sources from 3 authors.

    All 66 sources behind these findings
    1. Ep. #45, When Agents Outrun Human Review with Dotta of PaperclipDotta, cofounder of Paperclip (Paperclip, an open-source agent platform that has received about 5,000 community pull requests), Podcast episode transcript, Open Source Ready (Heavybit), published Oct 22, 2026
    2. Post by @caozillaMike Cao, runs the start-up Umami (Umami, open source web analytics company), X post, published Oct 20, 2026
    3. Riker: An Agent in My AgentNick Nisi, engineering manager on the DX / AI team at WorkOS (WorkOS DX and AI team; an internal agent manager hands work to coding agents and returns PRs with proof), Personal site post, nicknisi.com, published Oct 12, 2026
    4. Overview: What is Greptile?Greptile (Official), Product documentation, checked Oct 12, 2026
    5. Key featuresGreptile (Official), Product documentation, checked Oct 12, 2026
    6. Graph-based codebase contextGreptile (Official), Product documentation, checked Oct 12, 2026
    7. Memory and learningGreptile (Official), Product documentation, checked Oct 12, 2026
    8. Controlling nitpickinessGreptile (Official), Product documentation, checked Oct 12, 2026
    9. Review tiersGreptile (Official), Product documentation, checked Oct 12, 2026
    10. BillingGreptile (Official), Product documentation, checked Oct 12, 2026
    11. PricingGreptile (Official), Product documentation, checked Oct 12, 2026
    12. Open source programGreptile (Official), Product documentation, checked Oct 12, 2026
    13. Code providersGreptile (Official), Product documentation, checked Oct 12, 2026
    14. Auto-approve PRsGreptile (Official), Product documentation, checked Oct 12, 2026
    15. Fix with your AgentGreptile (Official), Product documentation, checked Oct 12, 2026
    16. Greptile CLIGreptile (Official), Product documentation, checked Oct 12, 2026
    17. MCP overviewGreptile (Official), Product documentation, checked Oct 12, 2026
    18. Organization settingsGreptile (Official), Product documentation, checked Oct 12, 2026
    19. Deployment optionsGreptile (Official), Product documentation, checked Oct 12, 2026
    20. Security practicesGreptile (Official), Product documentation, checked Oct 12, 2026
    21. ChangelogGreptile (Official), Product documentation, checked Oct 12, 2026
    22. feat(dev-loop): add Greptile and Graphite (free plans) to the review ladder after CodeRabbitRepository owner (self-described: One-maintainer setup whose agent loop falls back across several hosted and local reviewers), GitHub issue, published Oct 11, 2026
    23. Don't approve while a bug finding is open, and re-check approvals after new commitsTorsten Dittmann, product architect at Appwrite (self-described: Appwrite's appwrite/appwrite pull requests, where the author compares a self-built reviewer, Hansi, with Tenki and Greptile), GitHub issue, published Oct 6, 2026
    24. Platform stack: Greptile findings left after five review roundsContributor to pomerium/agentops (self-described: Pomerium's agentops repository, a stack of 12 pull requests reviewed by Greptile), GitHub issue, published Oct 6, 2026
    25. Deep, Codebase-Wide Reviews That Catch More IssuesPiyush T., reviewer on G2, self-described, G2 review, syndicated on AWS Marketplace, published Oct 6, 2026
    26. chore: retire coderabbit reviews, skip greptile on skip-agent-reviewGeorges-Antoine Assi, engineer at PostHog (self-described: PostHog's main repository, where the company retired CodeRabbit for its in-house reviewer), GitHub pull request on PostHog/posthog, published Oct 6, 2026
    27. Deep Codebase Understanding with Repo Graph, Bug & Security DetectionRasmita J., reviewer on G2 (self-described: Own team's repositories), G2 review, syndicated on AWS Marketplace, published Sep 25, 2026
    28. A connected reviewer we own replaces GreptileGrimblaz, maintainer of tradecraft (self-described: Own repositories reviewed by Greptile and Codex), GitHub issue, published Sep 24, 2026
    29. Compare Qodo and Greptile reviews and pick oneViktor Farcic, maintainer of dot-agent-deck (self-described: Open source repository reviewed by Greptile and Qodo), GitHub issue, published Sep 23, 2026
    30. Seamless GitHub Integration with Hands-OffTuli D., reviewer on G2 (self-described: Test pull requests in Python and JavaScript), G2 review, syndicated on AWS Marketplace, published Sep 19, 2026
    31. Speeds Up Understanding Large Codebases, Though Questions Need Careful PhrasingIzzy H., reviewer on G2 (self-described: Large unfamiliar codebase), G2 review, syndicated on AWS Marketplace, published Sep 15, 2026
    32. Owner governance: replace Greptile with Independent Agent ReviewOwner of Youtab B.V.'s Youtab-Agent-Runtime repository (self-described: Youtab B.V., an agent runtime repository with required merge gates), GitHub issue, published Sep 10, 2026
    33. [Technical] Code reviews at Spark: What we changed after AILuca Marturana, Chief Architect at Spark (self-described: Spark, early-stage energy software company; agents write most code, Claude Code style agent beside Greptile), Company blog post, SparkNotes (Spark), published Sep 9, 2026
    34. Post by @shcallawaySherwood, founder of Sazabi (X profile) (self-described: Sazabi engineering team), X post, published Sep 8, 2026
    35. Measure Greptile against the code-review skill: per-PR diff of findingsacharlop, whocards platform maintainer (self-described: whocards platform repository, free tier), GitHub issue, published Sep 7, 2026
    36. Comment on: Ask HN: Who wants to be hired? (September 2026)Senior Rails engineer, Levelbrook Consulting (HN profile) (self-described: Contract work on a large production codebase), Hacker News comment, published Sep 2, 2026
    37. Context-aware AI code reviews that actually understand the whole repositoryKirpalsinh R., reviewer on G2 (self-described: Team use), G2 review, syndicated on AWS Marketplace, published Sep 2, 2026
    38. Detailed PR Feedback, Fast Analysis, and Smooth GitHub IntegrationPrateek M., reviewer on G2 (self-described: Team use over a long period), G2 review, syndicated on AWS Marketplace, published Aug 27, 2026
    39. fix(ci): Dependabot PRs are unmergeable without admin because Greptile posts no required checkMaintainer of deftai/directive (self-described: deftai organization with a ruleset that requires a Greptile Review check on the default branch), GitHub issue, published Aug 24, 2026
    40. 3 things top 1% teams do differentlyAndrew Churchill, CTO of Weave (Weave engineering team; three AI reviewers on every PR, about 5% of PRs get a human review), Newsletter guest post, Manager.dev, published Jul 28, 2026
    41. greptile-gate: greptile-await.sh false TIMED_OUT, Greptile edits its summary comment in place (createdAt stays pre-trigger)Maintainer of Brite Nites' Claude Code plugins (self-described: Brite Nites, a merge gate that waits for Greptile's score on the brite-data-platform repository), GitHub issue, published Jul 8, 2026
    42. Comment on: Reviews have become expensive, rewrites have become cheapm463, Hacker News commenter, self-described, Hacker News comment, published Jun 16, 2026
    43. Comment on a Hacker News thread about human effort in pull requestsElFitz, senior software engineer at Purchasely (HN profile) (self-described: Purchasely, multi-model AI review on every pull request), Hacker News comment, published Jun 12, 2026
    44. Comment on: Ask HN: Is Anybody Using Codex?3dsnano, Hacker News commenter (self-described: Solo workflow with Claude Code and Codex), Hacker News comment, published Apr 30, 2026
    45. Comment on: Small models also found the vulnerabilities that Mythos foundepistasis, Hacker News commenter, self-described, Hacker News comment, published Apr 11, 2026
    46. Comment on: Slop is not necessarily the futurekusokurae, Hacker News commenter, self-described, Hacker News comment, published Apr 1, 2026
    47. Comment on: Code Review for Claude Coderaflueder, Hacker News commenter (self-described: Solo project), Hacker News comment, published Mar 9, 2026
    48. Comment on: Orchestrate teams of Claude Code sessionsNik Cubrilovic, developer (HN profile) (self-described: Own projects, Opus for coding and Codex for review), Hacker News comment, published Feb 5, 2026
    49. Post by @kyeburchardKye, builds agents at Falconer (X profile) (self-described: Single pull request comparison), X post, published Jan 30, 2026
    50. Comment on: There is an AI code review bubbleiblaine, data engineer (HN profile) (self-described: Data engineer), Hacker News comment, published Jan 27, 2026
    51. Comment on: There is an AI code review bubblesebra, Hacker News commenter (self-described: Employer team, Python codebase, previously on CodeRabbit), Hacker News comment, published Jan 27, 2026
    52. Comment on: There is an AI code review bubblemohsen1, Hacker News commenter (self-described: Own repositories), Hacker News comment, published Jan 26, 2026
    53. Comment on: There is an AI code review bubbledisillusionist, Hacker News commenter (self-described: Employer engineering team, evaluating against Cursor), Hacker News comment, published Jan 26, 2026
    54. Comment on: There is an AI code review bubbleseanmccann, Hacker News commenter (self-described: Own code written with Claude Code), Hacker News comment, published Jan 26, 2026
    55. Comment on: There is an AI code review bubblekaishin, Hacker News commenter (self-described: Employer team), Hacker News comment, published Jan 26, 2026
    56. Full Story of Brex’s AI Hail MaryJames Reggio, CTO at Brex (Brex, about 300 engineers; traditional linters plus Greptile for agentic code review), Podcast episode show notes and transcript, Latent Space, published Jan 17, 2026
    57. Comment on: The creator of Claude Code's Claude setupej88, Hacker News commenter (self-described: Solo, 3 to 5 parallel Claude Code sessions), Hacker News comment, published Jan 7, 2026
    58. Post by @michael_chomskyMichael, member of technical staff at context.dev (X profile) (self-described: Own projects on a Bun, oRPC and better-auth stack; free access from both vendors), X post, published Dec 22, 2025
    59. Comment on: Codemaps: Understand Code, Before You Vibe ItCSMastermind, Hacker News commenter (self-described: Team that compared several review bots), Hacker News comment, published Nov 5, 2025
    60. Comment on: Potential issues in curl found using AI assisted toolsAlex Chantavy, open source maintainer (HN profile) (self-described: Maintains an open source project), Hacker News comment, published Oct 2, 2025
    61. Comment on: AI-generated "workslop" is destroying productivity?bwfan123, Hacker News commenter, self-described, Hacker News comment, published Sep 22, 2025
    62. Comment on: AI tooling must be disclosed for contributionsOvais Tariq, co-founder and CEO at Tigris Data (HN profile) (self-described: Tigris Data), Hacker News comment, published Aug 21, 2025
    63. Comment on: Why Greptile just does code reviews and doesn't also generate codeJames Hartig, software engineer at Admiral (HN profile) (self-described: Admiral, engineering team), Hacker News comment, published Aug 4, 2025
    64. Greptile -- AI Pull Request Reviews in 2025Emily T. Burak, developer (self-described: Solace, a personal Go and HTMX project; Greptile with default settings beside GitHub Copilot and Cursor), Personal blog post, published May 25, 2025
    65. Comment on: Claude Integrationsclintonb, Hacker News commenter (self-described: Greptile comments on bug tickets in Linear), Hacker News comment, published May 2, 2025
    66. Comment on: Kilo Code: Speedrunning open source coding AI999900000999, Hacker News commenter (self-described: Feeding Greptile a Haxe framework repository), Hacker News comment, published Mar 26, 2025

    Mar 2025 to Oct 2026.

    48 sources from practitioners, by month and sentiment.

    The full readExplore all Greptile Signals4 praise themes, 4 concerns, 1 point of agreement and 2 disagreements across all three groups, with all 66 sources open to inspect.

    Where It Fits

    Where does Greptile belong in the stack?

    Open Where It Fits: the Stack Field and the full map

    Greptile is Core across Validate, supports Build and Ship, and touches Plan and Operate.

    Business area
    Product & Engineering
    Function
    Software Development
    Operating stack
    Agentic Software Development
    Primary role
    AI Code Review Agent
    Greptile beside the rest of its stackThe Stack Field, every product on every layer
    FootprintGreptile across the Agentic Software Development stack. Bar height is its role on each layer.

    1. Plan

      1. Work planning: Touches

        Work planning, touches. With the Jira or Linear integration, Greptile checks a pull request against its linked issue. Planning stays in the issue tracker.

    2. Build

      1. Codebase context: Supports

        Codebase context, supports. Each repository is indexed into a graph and a knowledge base, which Greptile's MCP server exposes to coding agents for review work.

      2. Coding: Touches

        Coding, touches. Review comments carry suggested fixes, and Fix with your Agent sends them to Claude Code, Codex, Conductor, Cursor, or Devin to apply.

      3. Agent runtime: Touches

        Agent runtime, touches. The Greptile CLI and the Claude Code and Codex plugins bring reviews into the terminal and into agents. Engineers steer the work in other tools.

      4. Tool execution: Supports

        Tool execution, supports. Through Greptile's MCP server, coding agents fetch review comments, trigger reviews, search past feedback, and read the knowledge base.

    3. Validate

      1. Debugging: Touches

        Debugging, touches. Greptile finds bugs before merge. Failures in running systems stay with debugging and observability tools.

      2. Testing: Supports

        Testing, supports. T-Rex, in beta, writes targeted tests for a pull request and runs them in a sandbox. It attaches logs and traces to failed findings.

      3. Code review: Core

        Code review, core. This is Greptile's primary role. It reviews every pull request with full repository context, at a strictness and tier the team sets.

      4. Application security: Supports

        Application security, supports. Reviews flag security issues by default, and rules can run deeper tiers on pull requests labeled security. Greptile is not a dedicated security scanner.

    4. Ship

      1. Source control: Supports

        Source control, supports. Reviews post on GitHub, GitLab, Bitbucket, and Gitea pull requests, and auto-approve, in beta, can approve low-risk ones. The code host stays the system of record.

      2. CI/CD: Touches

        CI/CD, touches. The Greptile CLI runs reviews in CI with an API key. Building and deploying stay with the delivery platform.

    5. Operate

      1. Observability: Touches

        Observability, touches. Analytics track review activity and bugs caught. Greptile does not monitor production systems.

    Market mapThe layers Greptile covers, and who else owns each one.

    1. Code review & quality

      CoreValidate

    2. Codebase context & intelligence

      SupportsBuild

    3. Tool execution & integrations

      SupportsBuild

    4. Testing & verification

      SupportsValidate

      • Playwright
      • Cypress

    5. Application security

      SupportsValidate

      • Snyk
      • Semgrep

    7 more layers on the full map. Core, Supports and Touches describe coverage. They say nothing about quality.

    More on Greptile

    1. GILD ReviewGILD's evaluation, with the buying facts
    2. Research1 piece
    3. Guides4 guides
    4. Where It FitsOn the Agentic Software Development stack
    5. Related Products6 products