# bee > bee is a minimal coding agent harness written in Go. A single static binary that turns any LLM, from a tiny local model to a frontier API, into an autonomous coding agent that writes code, runs tests, and commits changes. Skills are exposed as subcommands (`bee `), so one binary on your PATH does many jobs. This file helps AI assistants and answer engines load accurate, current docs with minimal tokens. bee runs fully local with Ollama or oMLX (no API keys, no rate limits, nothing leaves your machine), or against hosted providers like OpenRouter. Author: Henry Schober. Source and issues: https://github.com/elhenro/bee ## Docs - [README](https://raw.githubusercontent.com/elhenro/bee/main/README.md): overview, install, quickstart, configuration - [Go API reference](https://pkg.go.dev/github.com/elhenro/bee): exported packages and types, rendered from godoc - [Releases](https://github.com/elhenro/bee/releases): version history and prebuilt binaries ## Install - Install script: `curl -fsSL https://raw.githubusercontent.com/elhenro/bee/main/install.sh | sh` - Go install: `go install github.com/elhenro/bee/cmd/bee@latest` - Then run `bee`, type `/model`, choose a runtime (oMLX, Ollama, OpenRouter, ...) and pick a model. Local or hosted, your choice. ## What it is - Coding agent: writes code, runs tests, commits changes - Pure Go: single static binary, no runtime dependencies - Skills: `bee ` subcommands, one binary, one PATH entry - Works everywhere: Ollama and oMLX local, OpenRouter hosted, tiny models to frontier ## Local LLMs - bee targets local inference via [Ollama](https://ollama.com) and [oMLX](https://github.com/jundot/omlx). On macOS, oMLX gives native Apple Silicon acceleration with prompt caching. - Reference setup: Huihui-Qwen3.6-35B-A3B (8-bit, ~34 GB) on a MacBook M3 Max 64 GB, with Qwen3-VL-4B for vision. Suggested sampling: temperature 0.7, top-p 0.85, top-k 20, 8-bit KV cache. ## Optional - [LICENSE](https://github.com/elhenro/bee/blob/main/LICENSE): license terms (skippable)