octo-agent docs

octo is an MIT-licensed, single Go binary AI agent. It does what the big coding agents do — skills, CLI / Web / phone-IM, browser control, an OS-level sandbox — but as an open, self-contained binary you fully own, on any model (DeepSeek, Kimi, Anthropic, OpenAI, or anything compatible), with the server and your data staying on your own machine.

curl -fsSL https://octo-agent.dev/install.sh | sh     # single binary — no Node / Ruby / Python
octo config                                            # pick a provider, paste a key
octo "Add a --json flag to 'octo config show' and run the tests"

octo vs. the field

	octo-agent	Claude Code	Codex CLI	Hermes Agent
License / cost	MIT, free	proprietary, subscription	open source (Rust), OpenAI plan	MIT, free
Runtime	single Go binary, zero deps	native install	native Rust binary	Python (uv)
Models	Anthropic + OpenAI protocols, any compatible endpoint	Anthropic-first	OpenAI	multi-provider
Deployment	fully self-hosted	Anthropic-managed	OpenAI-hosted app + local CLI	self-hosted
Skills	Claude Code-compatible SKILL.md	native (origin format)	AGENTS.md / plugins	own format
Chat platforms	6 built in (WeChat, Feishu, DingTalk, WeCom, Discord, Telegram)	none built in	Slack / GitHub / Linear integrations	~20 platforms

Details per each project’s public docs. The difference is openness, self-hosting, and model freedom — not a feature checklist; octo covers the same ground as the closed alternatives.

Where to go next

Getting Started

Install octo and run your first task in under five minutes. Start with Install.

Guides

Task-first walkthroughs — MCP, skills, sandboxing, sub-agents, chat bridges. Start with Connect MCP servers.

Reference

Every CLI flag, config key, and tool, in one searchable set of tables. Start with the CLI reference.

Architecture

How the agent loop, providers, and tools fit together — for contributors. Start with System layers.