Hunkyard

Local-first AI code review for Git. Hunkyard captures a frozen snapshot of your changes, runs specialist reviewers against it with the model you choose, and gives you a browser workspace to triage findings, discuss them and approve fixes.

• Local by default. Works with Ollama, LM Studio, llama.cpp or any OpenAI-compatible server on your machine. Hosted providers (OpenAI, DeepSeek, AgentRouter) need explicit per-repository consent, and there is never a silent cloud fallback.
• Any Git scope. Review staged changes (including partial staging), unstaged or working-tree changes, a branch, a commit range or a GitHub pull request.
• Specialist reviewers. Correctness, security, frontend/accessibility, architecture/performance and testing, with an optional verifier pass.
• Respects your project's instructions. Reads AGENTS.md, CLAUDE.md, Copilot, Cursor and Qodo rules alongside Hunkyard's own config.
• Nothing changes without you. Hunkyard never stages, commits or pushes. Fixes are applied only after you preview and approve the exact change.
• Persistent history. Sessions, runs and decisions are stored locally in SQLite and survive restarts.

Requirements

• Node.js 22.22 or later (22 and 24 LTS are tested)
• Git
• A model: a local runtime such as Ollama, or an API key for a hosted provider

Install

npm install --global @hunkyard/cli

Or run it once without installing: npx @hunkyard/cli.

Usage

Point Hunkyard at a local model (Ollama here; use a model you have installed):

hunkyard config --provider ollama --endpoint http://127.0.0.1:11434 --model qwen2.5-coder:7b --privacy local-only
hunkyard doctor

For LM Studio, llama.cpp or another OpenAI-compatible server, use --provider lm-studio, llama-cpp or compatible with its --endpoint (for example http://127.0.0.1:1234/v1 or http://127.0.0.1:8080/v1). Hosted models are covered below.

Then, from any Git project:

hunkyard

This opens the browser workspace. Pick a scope and start a review. Keep the terminal open, and press Ctrl+C to stop.

You can also review from the command line:

hunkyard review --staged
hunkyard review --branch main --profile standard
hunkyard sessions

Run hunkyard --help for all commands and options.

Hosted models

Set the provider's API key in your environment (OPENAI_API_KEY, DEEPSEEK_API_KEY or AGENTROUTER_API_KEY), then allow cloud use for this repository:

hunkyard config --provider deepseek --model deepseek-flash --privacy cloud-allowed --allow-cloud

Revoke it any time with hunkyard config --revoke-cloud.

GitHub pull requests

With the GitHub CLI signed in (gh auth login):

hunkyard config --privacy cloud-allowed
hunkyard pr review 123 --allow-github

Publishing a summary comment back to the PR is a separate, confirmed step in the workspace.

VS Code

See the extension README.

Development

Requires pnpm 10.

corepack enable
pnpm install --frozen-lockfile
pnpm exec playwright install chromium
pnpm build
HUNKYARD_REPO=/path/to/a/repo pnpm dev   # API + UI on http://127.0.0.1:5173
pnpm check                                # lint, typecheck, tests, build and e2e

Commits follow Conventional Commits.

License

MIT