The independent reviewer for every pull request, whether Codex, Claude Code, Cursor or a person wrote it.
Free plan · 2-click install · No credit card required
Open a PR and HeyDeer reviews it. Click Autopilot, and HeyDeer fixes what it found until the PR is ready to merge.
Research: HeyDeer finds 78% more issues than GitHub Copilot on the same pull requests
When different models write the code, how many real issues does HeyDeer find?
A typo like 192.168.1 silently excludes a different address.
One network blip makes visitors re-enter the link’s password.
A saved IP range is accepted but never excludes anyone.
Anyone with a share link can read the excluded office IPs.
A network error tells visitors their link has expired.
Open shared dashboards reload every five seconds.
Protected links never unlock, and the password ends up in the URL.
Turning on password protection without a password leaves the link public.
Autopilot clears the routine work on each pull request, and the walkthrough shows you what still needs an engineer. For anything else, mention @heydeer.
HeyDeer handles the routine
Hand a pull request to Autopilot. HeyDeer fixes findings, failing checks and merge conflicts, pushes to the branch and reviews each change, round after round, until the pull request is ready to merge or needs you.
Read the docsYou handle the hard parts
Every review comes with a walkthrough of the change, group by group: what to read first, what to skim, and a diagram where parts connect.
Read the docsAnything else, just ask
Mention @heydeer on a pull request or issue to ask a question or hand off a job, like implementing an issue or fixing a finding. HeyDeer answers where you asked.
Read the docsHeyDeer follows your REVIEW.md, AGENTS.md and repository checks, and every comment cites the line behind it.
Coding agents follow fix requests even when they can tell they’re wrong (Jaiswal et al., 2026). So HeyDeer has a different model check each potential issue against the code, and posts only the ones that hold up.
99.0%Highest precision on ReviewBench GitHub’s open benchmark for AI code review+198 linesacross 8 files
The refund bug still ships.
+23 linesacross 2 files
Too much or too little? Set the feedback scope.
More answers in the docs and on the pricing page.
Start on the Free plan, with 1,000 credits in your first month and 300 every month after that. Starter is $9 a month and Pro is $39 a month. There are no seat fees, so adding developers doesn’t change the price.
Claude Code Review, Codex, Copilot and Cursor Bugbot come with a coding agent. HeyDeer is independent: it reviews every pull request the same way, whether Codex, Claude Code, Cursor or a person wrote it, and a different model checks each potential issue before it’s posted.
A different model checks each potential issue against the code, and only findings that hold up reach your pull request. Feedback scope lets you choose Concise, Balanced or Detailed for a workspace or a single repository.
Yes. HeyDeer reads REVIEW.md and AGENTS.md in your repository, runs the repository checks in .agents/checks/, and follows the review instructions you save in HeyDeer. Each comment cites the line it comes from.
Standard suits most pull requests. Deep follows the same steps but spends more time and credits on complex or sensitive changes. Set a default per workspace or repository, or pick Deep for one risky pull request.
No. Autopilot works a pull request until it’s ready to merge, and you merge it. HeyDeer merges only when someone with write access asks it to in a mention, and your branch rules apply as they do to anyone.
GitHub pull requests in public and private repositories. Reviews of public repositories are free up to a fair-use limit.
Start free with 1,000 credits in your first month, then 300 each month. No credit card required.
Get started free
