by Sonar
AI code review agent that diagnoses CI failures and pushes fixes to the pull request, free for open source and supporting 4 code hosts.
Pick Gitar if your team merges many AI-written pull requests and wants CI failures fixed, not just flagged. It replaces a review bot plus manual log reading. Teams above 50 users need Enterprise, and an editor-only developer gets little from it. The trial runs 14 days with no card.
It pushes a verified fix to the pull request and loops until CI passes, instead of stopping at a comment.
Tight hooks into code hosts and CI systems keep teams on it once rules and history exist, but any rival that adds CI-aware fixes narrows the gap.
Main trade-off
Pricing is per user and the first two plans stop at 50 users, so larger teams are pushed to Enterprise.
| Plan | Monthly price | What it includes |
|---|---|---|
| Open source | Free | Free for public repositories on GitHub or GitLab, Needs an OSI-approved licence such as MIT, Apache-2.0 or GPL, Same features as the Pro plan |
| Core | $20/mo | Annual billing; $25 per user on monthly billing, Up to 50 users, unlimited public and private repositories, High-signal code review and automatic PR summaries, CI failure diagnosis on GitHub Actions and GitLab Pipelines, On-demand fixes and an interactive agent on pull requests, Verification analytics and risk triage |
| Pro | $40/mo | Annual billing; $50 per user on monthly billing, Everything in Core, Auto-approve and merge blocking under your rules, Autonomous fix loop that commits on green, CI analysis for CircleCI, Buildkite and Bitrise, Slack, Linear and Jira integrations |
| Enterprise | Custom | Custom pricing, unlimited users, Everything in Pro, Self-hosted code hosting and bring your own LLM API key, SSO and SAML, audit logs, API access, Cross-repo analysis, Agentic Security and dedicated support |
From $20 per user a month when paid yearly; monthly billing adds $5 to $10. Pro is $40 yearly. Enterprise is custom, open source projects get the Pro feature set free, and a 14-day Pro trial needs no credit card.
Gitar is an AI code review and code change verification agent for GitHub, GitLab, Bitbucket and Azure DevOps. It reviews pull requests, diagnoses failing CI and commits fixes until the build passes. Sonar, the SonarQube maker, acquired it on 21 May 2026 and keeps it available as a standalone product.
High-signal code review
Reviews each pull request against your team's conventions and posts one consolidated comment, with inline notes only where they matter.
CI failure diagnosis
Finds the root cause of a failing build and separates real breakage from flaky tests and infrastructure noise on GitHub Actions and GitLab Pipelines.
Autonomous fix loop
On Pro, Gitar commits a fix and keeps iterating until CI passes, working only inside the rules you define.
Risk triage
Rates each whole pull request Low, Medium or High risk, even when the review finds no defects, and feeds that rating into approval and merge rules.
Cross-repo and cross-PR checks
Looks for related open pull requests or other repositories that a change could break, for example one PR removing an API field another still reads.
Custom checks in plain language
Teams write automations and policy checks in natural language and keep them in a .gitar folder in the repository.
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| $20/mo | Yes | Web | 3.7 | |
| $24/mo | Yes | Web · Mac · Windows | -- | |
| -- | Yes | Web | 3.9 | |
| $30/mo | Trial | Web · Mac · Windows | -- |
Proprietary
Core is $20 per user a month on annual billing or $25 monthly. Pro is $40 per user a month annual or $50 monthly. Enterprise is custom. Core and Pro cover up to 50 users and unlimited repositories. Monthly billing adds $5 to $10 per user compared with annual.
Maintainers of open source projects hosted publicly on GitHub or GitLab, under an OSI-approved licence, get the Pro feature set at no cost. Everyone else gets a 14-day free trial of Pro, with no credit card needed to start. After the trial a paid plan is required.
The docs list GitHub, GitLab, Bitbucket and Azure DevOps, with AWS CodeCommit in preview for Enterprise organisations on request. CI diagnosis covers GitHub Actions and GitLab Pipelines on Core, and adds CircleCI, Buildkite and Bitrise on Pro.
According to the pricing page FAQ, Gitar deletes code and data once an agent run completes, and neither it nor its LLM providers train on customer repositories. Enterprise customers can route all model calls through their own Anthropic API key.
Yes. Sonar's 21 May 2026 press release says Gitar stays available as a standalone product with no impact on existing customers, and can also be bought together with SonarQube. The founders joined Sonar to keep leading its development.
Gitar, at gitar.ai, is a code change verification agent that works inside pull requests across the major code hosts. It reviews a change, separates real CI breakage from flaky tests and infrastructure noise, and can push a fix to the branch. The company launched from stealth in 2026 and was bought by Sonar, the maker of SonarQube, on 21 May 2026. The two founders, Ali-Reza Adl-Tabatabai and Gautam Korlam, joined Sonar to keep leading the product, and Sonar says Gitar stays available as a standalone product.
What sets it apart from a plain review bot is the loop after the comment. Gitar reads the failing build log, proposes a patch, commits it, and on the Pro plan keeps iterating until CI passes. It runs on a frontier language model from Anthropic; the vendor's comparison page names Claude Opus 4.8 as the current model, and Enterprise teams can route calls through their own Anthropic API key.
Sonar describes Gitar as built for the stage after code is generated, so it suits engineering teams reviewing many AI-written pull requests. Platform engineers get review and CI dashboards, and team leads can set written rules for what Gitar may fix, what it must escalate and what blocks a merge. It also suits open source maintainers, since public repositories under an OSI-approved licence get it free.
It is a weaker fit if you want a reviewer that lives in your editor or one tied to a single code host. CodeRabbit and Greptile are review-first tools with their own codebase-indexing approach, and Qodo leans toward test generation alongside review. Cursor ships BugBot, which the vendor lists as a rival, and reviews inside the editor. The first two plans cap team size, so a larger company needs the Enterprise plan. Gitar sits in the AI coding assistants list, and the best AI code review tools guide compares it with its closest rivals. The product now belongs to a bigger company, so its roadmap will follow SonarQube integration over time.
Our call: choose Gitar if your team runs GitHub or GitLab CI and wants failing builds fixed, not just flagged. Choose CodeRabbit or Greptile if you only want review, and skip Gitar if you work alone inside an editor.
Product facts on this page (pricing, plans, integrations, compliance) are read from Sonar's own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.