The version comparison is the tool that changes a workflow: point it at before and after, and it tells you whether the change introduced a regression and whether it should be merged.
An MCP server that sends code to a Llama model hosted on Groq for review and returns structured findings — severity, line, title, description and a suggested fix — alongside a score from 0 to 100. It reviews snippets or whole files, compares versions, generates tests and builds reports.
- A strict review of a snippet: bugs, security issues, and a score from 0 to 100
- Findings as structured JSON with severity, line number, description and a fix
- Two versions compared, regressions detected, and a merge or request_changes recommendation
- A step-by-step explanation pitched at a junior, middle or senior audience
- Pytest, jest or go test generated — happy path, edge cases and security
- A whole file from disk analysed with parallel chunking
- An HTML report built from any analysis result
- Cache statistics, and the cache cleared
A recent Python and a free Groq API key in GROQ_API_KEY, obtainable without a card. Code you submit is sent to the Groq API to be reviewed.
One command plus a key — uvx mcp-code-sanitizer, then supply credentials
