Code execution via eval() on model output
OhMyBug caught it when we replayed the change that introduced it. In the code it lived ~7 months before the fix.
CVE-2026-35002 · advisory published June 30, 2026
Built for code your agents write. The result goes to your coding agent: it checks each finding, fixes the real medium or worse bugs and offers to record the real lows. There is no review inbox for people; you read a short list.
Pick /code-review if you want a quick review inside the Claude Code session you are already in, billed like the rest of that session. Pick OhMyBug if you want a bug hunt before you merge or push: from $1 per review that finds a bug on a pushed commit, $4 fast on any change, $0 when the code is clean (mode prices).
/code-review Each review uses plan usage or API billing
10 of 10 reviews use plan usage
OhMyBug Charged only when a review finds a bug
3 of 10 reviews charged
| Topic | OhMyBug | /code-review |
|---|---|---|
| You pay for | From $1 per review that finds a medium or worse bug on a pushed commit, $4 fast on any change; $0 when clean. See mode prices. | Your Claude plan's usage or API billing, for every review |
| Where it reviews | Fast covers any change your coding agent can see, with or without a pull request | Your branch, a pull request, a path, or a revision range, from a Claude Code session |
| Where it runs | Our cloud sandbox | Your own Claude Code session |
/code-review facts as of November 19, 2026. The sourced facts are in the full table.
OhMyBug #1 on ReviewBench · snapshot October 25, 2026
Each finding is checked by your own Claude or Codex, against your repository, before you act on it.
No public ReviewBench result for Claude Code /code-review as of November 19, 2026.
Grounded scores from review-bench.ai, pulled November 11, 2026.
| Reviewer | Found | Precision | Findings per change | Wait, typical | Wait, slowest 10% | Price per change |
|---|---|---|---|---|---|---|
| OhMyBug lite | 61.5% | 79.9% | 49 | 2.1 min | 2.6 min | $1 |
| OhMyBug snap | 57.7% | 82.7% | 45 | 56 s | 83 s | $2 |
| OhMyBug fast* | 76.5% | 82.5% | 61 | 2.3 min | 6.3 min | $4 |
| Claude Code /code-review (xhigh) | 42.9% | 89.4% | 13 | 7.0 min | 11.6 min | $1.55 API cost |
Data set: OMBench, our held-out set of 85 changes from public repositories with 742 known bugs. Every reviewer's findings were judged by the same model under the same rules. Found is the share of the known bugs a review reported. Precision takes the findings the judge could match to its answer list and gives the share of them that are real bugs. Wait is from sending the change to getting the result; 10% of changes waited longer than the second figure. Our price is charged only when a review finds a medium or worse bug, $0 when clean. The /code-review figure is its API cost per change at Anthropic's list price, median; on a Claude plan the review uses the plan's usage instead. How we count
* Measured on the review as delivered today.
/code-review: Claude Code 2.1.325, /code-review xhigh on Opus 5.5, run by us on November 18, 2026 from a fresh setup with no plugins, on a checkout of the base with the change staged. It could read the whole repository and use git, with no web access and no builds or tests. It used its default finding limit (at most 15 per change) and gives no severity. OhMyBug: lite measured on November 17, 2026, snap measured on November 16, 2026, fast measured on November 17, 2026.
eval() on model outputOhMyBug caught it when we replayed the change that introduced it. In the code it lived ~7 months before the fix.
CVE-2026-35002 · advisory published June 30, 2026
OhMyBug caught it when we replayed the change that introduced it. In the code it lived ~14 months before the fix.
GHSA-mchm-gjx6-7h55 · advisory published September 24, 2026
From the replays on our home page; advisories checked November 10, 2026.
A lite review of a pushed commit costs $1, a fast review of any change $4, when it finds a medium, high, or critical bug. Clean reviews, minor-only reviews, and findings your agent rejects cost $0. Your first bug-finding review is free. How billing works
Different on CI: CI billing
/code-review counts toward your plan's usage or API billing the same way the rest of the session does. Source
The hunt runs in our cloud. Each finding comes with a severity and a failure scenario. Your coding agent checks every one against your repository, sends the verdict back, and fixes the confirmed medium or worse bugs. It does not fix lows blindly or drop them: it offers to record each real one where the next review sees it. You approve the verdicts. How an agent should use OhMyBug
From $1 per review when it finds a bug on a pushed commit ($4 fast on any change; mode prices). A deep hunt runs only when you choose it, and the offer shows the price first: $12, or $24 on a very wide change. One bug or ten costs the same.
OhMyBug is ranked #1 on open ReviewBench as of November 11, 2026 (review-bench.ai). See real bugs it found this week in the Bugarium.
Claude Code /code-review facts as of November 19, 2026, quoted from its public pages
| Topic | OhMyBug | /code-review |
|---|---|---|
| Price | From $1 per review that finds a bug on a pushed commit; $0 when clean. Fast review $4 covers any change and costs only when it finds a medium or worse bug. Lite and snap prices and eligibility are on the modes page. A deep hunt runs only when you choose it: $12, or $24 on a very wide change. | "The command works on any plan. The review runs in your own session, so it counts toward your plan's usage or API billing the same way the rest of the session does." docs |
| Where it reviews | Any change your coding agent can see: a branch, uncommitted changes, recent commits, or a folder with no git. In CI, the OhMyBug CLI. | "A target can be a pull request number, a branch name, a file path, or a Git revision range." docs |
| Where the review runs | Our cloud sandbox. No self-hosting. What leaves your machine. What is kept and for how long. | Your own Claude Code session: "In a terminal session, the review runs as a background subagent with its own context window." docs |
It can be, if you would rather pay per review that finds a bug. As of November 19, 2026, Claude Code's docs say /code-review works on any plan and counts toward your plan's usage or API billing the same way the rest of the session does. An OhMyBug fast review costs $4 when it finds a medium or worse bug and $0 for a clean one; other modes are on the modes page. Some teams use both: a reviewer on every pull request, and a bug hunt before a risky merge.
Yes. OhMyBug has no seat fee and no monthly plan. You add credit, and a fast review costs $4 when it finds a medium, high, or critical bug. Your first bug-finding review is free, with no card.
No, per review. One bug or ten costs the same: $4 for a fast review (other modes). A clean review, or one with only minor findings, costs $0.
The review gives each finding a severity and a failure scenario. Your coding agent checks every finding in your repository and marks it REAL or NOT REAL, and you approve the verdicts. Findings it rejects cost nothing. On CI billing there is no confirm step. The full rules are in the FAQ.
No. Your coding agent can send a branch, uncommitted changes, your last commits, or a folder with no git. If you commit straight to main, ask it to hunt bugs in your last few commits.
Your diff and a snapshot of your repository, without secret-named files and ignored files that are not committed. A pushed commit we can read is read on GitHub instead when your agent or the CLI points us to it. Diff mode sends the diff and selected files without the snapshot, except for a deep review. Reviewers may ask for more files. Submitted diffs and files are removed from the database after the sandbox receives them; an uploaded snapshot is deleted by our regular cleanup after the review ends, and an extra run of the same review keeps its own copy until that run ends. Run transcripts are kept for 7 days after a completed review and 30 days after a failed one. Submissions are not used to train models. Details: How it works.
As of November 19, 2026, its docs say --max-findings reports up to a number you choose, or every finding, "in place of the review's usual limit". In our runs at effort xhigh it reported at most 15 findings per change.
Claude Code:
claude plugin marketplace add OhMyX-AI/ohmybug-plugin
claude plugin install bughunter@ohmybug
Codex (the update runs last):
codex plugin marketplace add OhMyX-AI/ohmybug-plugin
codex plugin add bughunter@ohmybug
codex mcp login ohmybug
codex update
Then say "hunt bugs" in your project. First bug-finding review free. No card.
Start free review →Claude Code /code-review and other product names on this page belong to their owners and are used only to identify those products. Facts about Claude Code /code-review are quoted from its public pages as of November 19, 2026 and may have changed since; follow the source links for the current terms.