Back to directory

brain0

A passive, offline-by-default tool that builds a decision graph linking every git commit to the coding-agent prompt that produced it, with drift detection, a DLP audit of what agents read, and signed provenance attestations.

brain0

brain0 answers not just what changed in a repository but why: it passively reads git history and the transcripts that coding agents such as Codex and Claude Code already write to disk, with no hooks, no agent cooperation and no code changes required, and links every commit, down to the individual function, to the agent intent behind it, with dated history and drift detection.

A single npx brain0 up infers the repo ID from the git remote, indexes commit history, ingests coding-agent sessions, and opens a local GUI at http://localhost:8787 with an explorable graph from repo to module, file and symbol; clicking a commit shows the prompts behind it, per-file diffs and a two-dimensional risk score rendered green to red. Beyond the GUI, brain0 today gives a morning triage of what agents did, riskiest first, brain0 report produces an accountability report, and brain0 query supports root-cause debugging over the graph. Node.js 20+ is the only hard requirement; it works fully offline with deterministic summaries and local feature-hash embeddings, and can optionally use Ollama or a hosted LLM for richer summaries and semantic search.

More in AI evals

See all 269
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • JevBench

    benchmarkheaven.com

    A reproducible benchmark from Benchmark Heaven that ranks Jev-class typed decision models on intelligence, calibration, speed, and cost, with a published harness, scoring rules, and results.

  • A co-evolutionary adversarial red-teaming tool for AI agents that uses quality-diversity evolutionary search to autonomously breed edge-case scenarios, testing whether an agent with tool access can be made to hallucinate, deadlock or leak funds.