Introducing Hawk: Your Local AI Coding Agent
We've been building AI-powered developer tools for the past year. This post introduces Hawk — we're developing it in the open on GitHub; there is no public install yet.
What is Hawk?
Hawk is an AI coding agent that lives in your terminal. It reads your codebase, understands context, writes code, fixes bugs, and refactors systems — all without sending a single line to someone else's server.
You bring your own API key (Claude, GPT, Gemini, Ollama, or any OpenRouter model), and Hawk handles the rest.
Why local-first?
Every other AI coding tool wants you to upload your code to their cloud. We think that's backwards. Your code is your most sensitive asset. It should never leave your machine unless you explicitly choose to share it.
Hawk runs 100% locally. The only network calls are to your chosen LLM provider, using your own API key.
What's included
Hawk is designed to ship as a single CLI with integrated support capabilities:
- Hawk — The AI coding agent itself
- Eyrie — Support capability repo for LLM provider runtime (11 gateways, caching, routing)
- Trace — Support capability repo for session capture and replay
- Tok — Support capability repo for token compression (60-90% savings)
- Yaad — Support capability repo for graph memory across sessions
- Inspect — Support capability repo for website verification and audit
- Sight — Support capability repo for AI-assisted code review
Getting started
We're building Hawk in the open — local-first, Docker by default, model-agnostic. Public install is not available yet; follow GitHub for progress. When it ships: /config for API keys and models.
Updated May 2026: Eyrie now supports 11 gateways. Hawk includes 12,147 community skills. Gitant and Lark are announced as standalone products in the GrayCode stack.