San is an open-source terminal agent runtime: one native Go binary, no Node.js or Python. Everything the model touches — prompt, tools, providers, extensions — stays yours to change.
curl -fsSL https://raw.githubusercontent.com/genai-io/san/main/install.sh | bash
Open the full-quality intro ↗ · 54s, no sound
Scripted turn, example repo · what a real San run looks like
Three properties — 小 small, 快 fast, 开 open — and San refuses to trade any one of them for the others. Three strokes, one 三.
What the harness sends before your first message stays small, so the window belongs to the task at hand. On disk it is one file that needs nothing else installed.
Native Go, so nothing boots before you type: no interpreter, no node_modules to read off disk. What you wait on is the model.
What the model sees and what it may do are both yours to write — and to replay afterwards.
Anthropic, OpenAI, Google, DeepSeek, Ollama and a dozen more providers; the search backend of your choice; skills, plugins, MCP servers, subagents and hooks, all unmodified.
Compose the system prompt, bundle it into a persona you can switch, hand autopilot a goal, set the strategy self-learning follows.
You choose how much San may do without asking, and subagents inherit that choice. The inspector replays a run exactly as the model saw it.
Against Claude Code v2.1.112 on Apple Silicon, same model (claude-sonnet-4-6), same tasks.
| Metric | San | Claude Code | Advantage |
|---|---|---|---|
| Download size | 12 MB | 63 MB + Node 112 MB | 5× smaller |
| Disk footprint | 38 MB | 175 MB | 4.6× smaller |
| Startup time | ~0.01s | ~0.20s | 20× faster |
| Startup memory | ~32 MB | ~189 MB | 5.8× less |
| Simple task | ~2.4s / 39 MB | ~10.4s / 286 MB | 4.3× faster |
| Tool-use task | ~3.3s / 39 MB | ~26.0s / 285 MB | 7.9× faster |
| Harness context | ~2.3k tokens | ~20.9k tokens | ~9× leaner |
Comparable feature sets — San runs Claude Code's skills, plugins and MCP servers unmodified — so the gap is client-side overhead, not capability. Method and how to reproduce →
WeChat for the Chinese community, Slack for everyone else.
关注公众号「极客外传」· 回复 san 或 三 入群
Scan, or join us on Slack