'Heimdall' scans your code for security flaws using Claude, Gemini, or any local AI you already have

Heimdall is a free, open-source that finds security in your codebase by routing your files to whichever local AI tools you already have installed — Claude Code, Gemini CLI, Codex, or Opencode. You point it at a source folder, it sends the files to the AI, collects the findings, and produces a clean report in JSON, Markdown, or SARIF format. You can run multiple AI backends in parallel; Claude and Gemini often catch different issues, so combining them gives broader coverage.

A built-in step ensures the same problem is shown only once, whether two AI tools flagged it or it appeared in a previous scan. Everything stays on your local machine — no code is sent to external servers and no separate API keys are needed. It works across languages including JavaScript, Python, Go, Java, Rust, C#, and PHP.

Running `heimdall web` opens a local dashboard at port 4040 where you can browse past scan results and .

Key points

  • Reuses local AI tools you already have — no extra API keys or cloud accounts needed
  • Code never leaves your machine, making it safe for sensitive projects
  • Running Claude and Gemini together catches more issues than either alone
  • removes repeated findings across tools and across scan runs
  • One-liner install via curl; web dashboard for browsing scan history at port 4040
Read original