Vet is a fast and local code review tool open-sourced by the Imbue team. It’s concise where others are verbose, and it catches more relevant issues.
Confidence scoring and deduplication trim findings to what's worth your attention.
A reusable GitHub Action posts focused review comments on every pull request.
Built-in checks for correctness, security, commit quality, and agent-specific failures.
Add, override, or replace any guide in a few lines of TOML.
Clear exit codes plus text, json, and github output formats.
Named TOML profiles lock in model, format, and enabled checks so the team stays in sync.
When you use an agent to write code, things go wrong in ways you can't see in the diff. It tells you tests passed when it never ran them. It hits a wall and swaps in fake data. I built Vet to catch what standard code review misses, by reading the conversation alongside the changes.

Vet reads your agent's conversation history alongside the diff, so it catches when the agent's actions don't match your request. It runs before you push, works with local models, has no telemetry, and is tuned for precision over recall.
No. Zero telemetry, no account. Requests go directly from your machine to whichever inference provider you choose. With a local model, your code never leaves your machine.
Anthropic, OpenAI, Gemini, or any OpenAI-compatible endpoint, including local models via Ollama or compatible runtimes. Use your own API key, or pass --agentic to use an existing Anthropic or OpenAI subscription.
Yes. Install Vet as an agent skill and your agent runs it automatically after writing code, surfacing issues to fix in the same session.
Vet started as imbue_verify inside Sculptor. It's now standalone, so it goes everywhere Sculptor doesn't. Sculptor still ships with Vet built in.
AGPL-3.0. Free for any use, commercial included. If you modify Vet and offer it as a hosted service, you share your changes back.