🖥️ Savant Code — 9 AI Agents That Audit Every Change Before It Touches Your Repo (Free · Local · Open-Source)

Tools & Scripts
,

AI-generated summary

:bullseye: What you get: A free, local-first multi-agent coding tool that audits and tests code changes before touching your repository.

:white_check_mark: Start here:

  • GitHub repo — source code, setup instructions, and architecture details.
  • npm package — install globally with npm install -g savant-code.
  • Detective and Verifier pipeline — read-only agents that spot bugs and audit fixes independently.

:toolbox: Also useful:

  • Checkpoint and rewind — rollback changes and conversation state easily.
  • HTML audit reports — self-contained logs tracking agent reasoning.

:warning: Watch out:

  • Test it on a small codebase first to confirm the verifier catches planted bugs before relying on it.
1

:red_circle: RED ▸ :green_circle: GREEN ▸ :magnifying_glass_tilted_left: AUDIT ▸ :wrench: FIX ▸ :white_check_mark: DONE

no code touches your repo until it clears all five — enforced by 9 agents, on your own machine

Most AI coders are one model guessing — they edit, hope, and fail confidently. Savant Code runs 9 specialized agents with separated duties that audit every change before it hits your files. Local-first (Ollama), Apache-2.0, free.

npm install -g savant-code
cd your-project && savant-code     # auto-detects a running Ollama

:light_bulb: The insight: code quality isn’t a model problem, it’s a harness problem — you don’t wait for GPT-6, you engineer a pipeline that catches mistakes structurally. Detective finds bugs with evidence → Forge fixes surgically → a separate Verifier audits → self-correct → done. No merge without an independent audit.

🔵 The 9 agents + the engineering (for the tinkerers)
Agent Role
:compass: Savant orchestrator — routes work, enforces protocol
:red_circle: Detective RED — finds bugs with evidence · read-only
:green_circle: Forge GREEN — implements the converged plan · write
:magnifying_glass_tilted_left: Verifier AUDIT — independent double-check afterwards
:brain::telescope::globe_with_meridians::card_index_dividers::writing_hand: Thinker · Scout · Researcher · Recorder · Scribe reason · explore · web-research · track · document
  • Tool-permission boundaries — each agent gets exactly the tools its role allows (allowlist-filtered), no more
  • Fail-closed streaming — malformed tool calls are rejected, never coerced
  • Checkpoint & /rewind — per-turn edit checkpoints (code / convo / both / fork)
  • Provider-flexible — Ollama · OpenRouter · NVIDIA NIM · any OpenAI-compatible API
  • GitHub MCP (read-only PR/issue/CI review) · DB helper (read-only default, LIMIT injection, SQL redaction) · self-contained HTML reports
  • TypeScript/Bun · ECHO Protocol v0.2.0 · Apache-2.0

Grab it → GitHub · npm

One model guesses. Nine agents cross-examine. Your repo only ever sees what survives the audit.

6 Likes
2

The Detective-Verifier split is the interesting idea here, more than the agent count. One model writing and checking its own work tends to approve its own mistakes, so a separate read-only check with its own evidence trail is a real improvement. I would still start small: run it on a tiny repo, look at the HTML report for one fix, and check whether the Verifier actually catches a planted bug before trusting it on anything bigger. The checkpoint and rewind feature matters just as much, because an audit is only useful if rolling back a bad change is cheap.

1 Like
3

I appreciate the response, as the creator of Savant Code, i suggest you take it for a spin. It has tons of connections and free model access :slight_smile: