> For the complete documentation index, see [llms.txt](https://help.swarmia.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.swarmia.com/features/ai-tools/review-agents.md).
# Review agents
See which AI agents are reviewing your pull requests, how much of your code they're covering, and how many findings they're leaving per PR.
## What is a review agent?
A review agent is an AI tool that automatically reviews pull requests — leaving comments, flagging potential issues, and suggesting improvements — without creating any code itself. Unlike [cloud agents](/features/ai-tools/cloud-agents.md), which author PRs from start to finish, review agents act as an additional reviewer on top of your existing pull request process.
### Built-in review agents
* [**GitHub Copilot code review**](https://docs.github.com/en/copilot/using-github-copilot/code-review/using-copilot-code-review) can review pull requests directly on GitHub and leave inline suggestions.
* [**Cursor Bugbot**](https://cursor.com/bugbot) scans pull requests for bugs and posts a summary to the PR description.
* [**Claude Code**](https://code.claude.com/docs/en/code-review) can review pull requests when triggered via GitHub Actions or mentioned in a PR.
* [**Codex code review**](https://developers.openai.com/codex/integrations/github) reviews pull requests on GitHub, either automatically or when you comment `@codex review`.
* [**CodeRabbit**](https://www.coderabbit.ai) reviews pull requests on GitHub and GitLab, leaving inline comments and a summary of its findings.
Don't see your tool? [Add it as a custom AI tool](/settings/integrations/ai-coding-tool-integrations/custom-ai-tool-integrations.md#link-contributors) to get the same detection.
## Understanding your review agent coverage
Swarmia automatically detects pull requests reviewed by any of the agents above, as well as any [custom AI tool](/settings/integrations/ai-coding-tool-integrations/custom-ai-tool-integrations.md#link-contributors) you've linked a contributor to, and shows them in the [**AI tools → Review agents**](https://app.swarmia.com/ai/reviews).
### Agent-reviewed PRs
The **Agent-reviewed PRs** chart shows the share of your pull requests that received at least one review agent comment over time. The stacked bars distinguish PRs reviewed by an agent from those that weren't, and the review ratio makes it easy to track adoption across your organization.
Use this chart to understand whether review agent usage is growing and to spot teams that haven't adopted them yet.
### Findings per PR
The **Findings per PR** chart shows the distribution of pull requests by the number of comments review agents left on them. "Findings" are individual review comments or suggestions from the agent — not just a summary at the top level.
A healthy distribution means agents are leaving substantive feedback rather than just passing PRs silently. If you see a spike at zero findings, it may mean agents are being triggered but not surfacing meaningful feedback.
### Tabs: Teams, pull requests, and review agents
The table below the charts has three tabs:
* **Teams**: Breaks down agent-reviewed PRs, coverage percentage, total findings, and findings per PR by team. Useful for comparing adoption across your organization.
* **Pull requests**: Lists individual PRs with the agents that reviewed them and how many findings each agent left. Use this together with the PR filters to inspect specific repositories or time ranges.
* **Review agents**: Shows per-agent stats — how many PRs each agent reviewed, their share of total PRs, and their average findings per PR.
## Frequently asked questions
### Which review agents does Swarmia support?
We have built-in support for the review agents listed [above](#built-in-review-agents). For any other review agent, or one you've built yourself, you can [add it as a custom AI tool and link its contributor](/settings/integrations/ai-coding-tool-integrations/custom-ai-tool-integrations.md#link-contributors). Swarmia then attributes its reviews to it with high confidence, the same way it recognizes the built-in agents.
### How does Swarmia detect review agents?
Swarmia identifies a review agent by matching the author of review comments against known email addresses and GitHub login patterns for the agents listed above, or against the contributor linked to a [custom AI tool](/settings/integrations/ai-coding-tool-integrations/custom-ai-tool-integrations.md#link-contributors). On GitLab, where bots don't have a GitHub login, CodeRabbit is matched by its display name instead.
Swarmia doesn't distinguish whether the agent reviews pull requests proactively or if someone manually requests a review through GitHub, Slack, or another remote workflow. If the agent posts a review comment on the pull request, Swarmia counts it as a review-agent review in either case.
This differs from someone using an agent locally to help review a pull request. If they post the feedback from their own account, Swarmia treats it as a human review, not a review-agent review.
CodeRabbit requires no setup on your end. Swarmia detects it from pull request reviews, comments, and authors, as well as commit authors and co-authors, as soon as it's active in one of your repositories.
### Why does CodeRabbit show up with zero findings on some reviewed PRs?
When CodeRabbit doesn't find anything to flag, it posts a summary comment instead of a formal review. Swarmia still counts the pull request as reviewed by CodeRabbit, but with zero findings, since there's no inline comment to count.
### Why is a Codex review missing from a pull request?
Swarmia counts a [Codex](/settings/integrations/ai-coding-tool-integrations/codex-integration.md) review when Codex posts a review with inline findings. When Codex doesn't find anything to flag, it reports that in a way Swarmia doesn't yet recognize as a review:
* **Automatic reviews** leave only a 👍 reaction on the pull request description.
* **Reviews you trigger with `@codex review`** leave a plain comment, such as *"Codex Review: Didn't find any major issues."*
Neither is a review in GitHub's data, so those pull requests don't count as reviewed by Codex. This means Codex coverage can look lower than it is, while findings per PR are unaffected. We're working on covering both cases.
### Why doesn't Swarmia show findings from my custom review agent?
For [custom review agents you configure](/settings/integrations/ai-coding-tool-integrations/custom-ai-tool-integrations.md#link-contributors) that leave reviews on GitHub, each (non-deleted) inline review comment counts as one finding. Swarmia doesn't count replies to another review comment.
Comments posted in the review summary or elsewhere on the pull request don't count as findings. For example, a review that lists every finding in one summary comment has zero findings in Swarmia.
### How are review agents different from cloud agents?
[Cloud agents](/features/ai-tools/cloud-agents.md) create pull requests end-to-end. Review agents only leave feedback on pull requests created by humans (or cloud agents). The two can overlap: a cloud agent PR can also be reviewed by a review agent.
Swarmia tracks them separately so you can understand each type of AI involvement in your workflow. For example, CodeRabbit is usually a review agent, but if it opens a pull request itself, such as to apply a fix it suggested, Swarmia attributes that PR to CodeRabbit's [cloud agent](/features/ai-tools/cloud-agents.md) mode instead.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://help.swarmia.com/features/ai-tools/review-agents.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
