A second pair of eyes catches what you stopped seeing hours ago. This tool gives any code snippet a structured review using your own OpenAI or Anthropic key — pick a language and a focus, and get findings ordered by severity with concrete fixes, all in your browser.
How it works
Choose a provider and model, paste your API key, drop in your code, select the language, and pick a review focus: everything, bugs, style, performance, or security. The tool builds a prompt that frames the model as a senior engineer in that language, tells it exactly what to look for, and asks for findings ordered most-serious-first — each with the relevant construct, why it matters, and a suggested fix. It sends one direct request to the provider and returns the review in Markdown.
For Anthropic, the request includes the official direct-browser-access header so it works straight from the page.
What each review focus looks for
Bugs — the focus targets logic errors, incorrect conditionals, off-by-one issues, missed null checks, incorrect type assumptions, and cases where the code does not handle edge inputs or failure modes that are plausible in production. The prompt asks the model to reason about what the code is supposed to do, not just what it literally does.
Style — naming clarity, function and module size, adherence to idiomatic patterns for the language, unnecessary complexity, and maintainability concerns. Style findings have lower priority than bugs but matter for long-lived codebases where reading speed and refactoring ease compound.
Performance — unnecessary work in hot paths, O(n²) or worse algorithmic choices where a simpler approach exists, redundant database or network calls, blocking operations that should be asynchronous, and memory patterns that could cause leaks or excessive allocation.
Security — SQL or command injection risks, unsafe handling of user-supplied input, hardcoded credentials or secrets, insufficient authentication or authorisation checks, insecure direct object references, and unsafe deserialization. This focus is most useful as a complement to SAST tooling rather than a replacement, and it is worth running on any code path that touches user input or external data.
Everything — a balanced pass across all four categories, with findings ranked so the most serious (typically security and bugs) appear first regardless of category.
Getting the most useful findings
Snippet size matters. Sending a 50-line function yields sharper analysis than sending a 500-line file, because the model can reason about the logic in detail rather than summarising at a high level. For large files, split them and review the most critical functions first.
Language selection tells the model which idioms and pitfalls apply. A Python review looks for different patterns than a Go review — type hints, list comprehension misuse, and GIL concerns in Python; goroutine leaks, nil handling, and defer ordering in Go. Selecting “Other” still works but disables language-specific guidance.
Tips
- Review focused snippets rather than whole files — tighter context yields sharper findings.
- Run a Security pass on anything that touches user input, auth, or external data.
- Always verify a finding before acting on it; AI reviews are a prioritised to-check list, not a verdict.
- Cross-check security findings against your actual data flow before fixing — the model sees only the snippet, not the broader context of where input originates.