Hamr

Local-first coding agent. Built for local models on consumer GPUs.

Quick Start

npm install -g @skaft/hamr
hamr

Start your inference server (Relay, llama.cpp, LM Studio, Ollama, vLLM) at its default port, then run hamr. From the TUI: /login → "Use a custom/self-hosted endpoint" to configure.

Why Hamr

• Relay-first — recommended inference gateway: model lifecycle, prefix-cache pre-warming
• Local-first — designed for your GPU, not a cloud bill
• Pi SDK — embed the agent in your own tools
• OpenAI-compatible — works with any /chat/completions endpoint

Docs

• Getting Started
• Configuration
• Providers & Endpoints
• Relay Setup
• Commands & TUI
• Architecture