OpenFang: The Complete Guide to the Open-Source Agent Operating System
Most agent frameworks are Python wrappers around an LLM API — they wait for you to type something. OpenFang is different. It's a full Agent Operating System built from scratch in Rust: 137,000+ lines of code, 14 crates, a single ~32MB binary. Your agents don't wait — they run autonomously, on schedules, 24/7, delivering results to your dashboard. With 12,000+ GitHub stars in its first weeks, OpenFang is rewriting the rules of what an agent platform can be.
What Is OpenFang?
Not a chatbot framework. Not a multi-agent orchestrator. An Agent Operating System — a complete runtime where autonomous agents work for you: researching competitors, generating leads, managing social media, monitoring targets, and building knowledge graphs. All on schedule, all without prompting.
- Language: Rust
- License: Apache 2.0
- Stars: 12,500+ ⭐
- Forks: 1,400+
- Releases: 51
- Binary Size: ~32MB
- Cold Start: 180ms
- Idle Memory: 40MB
- Install:
curl -fsSL https://openfang.sh/install | sh
The Core Innovation: Hands
"Traditional agents wait for you to type. Hands work for you."
Hands are pre-built autonomous capability packages that run independently, on schedules, without prompting. This isn't a chatbot — this is an agent that wakes up at 6 AM, researches your competitors, builds a knowledge graph, scores findings, and delivers a report to your Telegram before you've had coffee.
Each Hand bundles:
- HAND.toml — Manifest declaring tools, settings, requirements, and dashboard metrics
- System Prompt — Multi-phase operational playbook (500+ word expert procedures)
- SKILL.md — Domain expertise injected into context at runtime
- Guardrails — Approval gates for sensitive actions
The 7 Bundled Hands
| Hand | What It Does |
|---|---|
| Researcher | Market research, competitor analysis, knowledge graph building |
| Lead | Lead generation on daily schedules |
| Social Media | Social media management and monitoring |
| Browser | Web automation with approval gates for purchases |
| Coder | Code generation and development tasks |
| Data | Data analysis and reporting |
| Ops | Operations and infrastructure management |
openfang hand activate researcher # Starts working immediately
openfang hand status researcher # Check progress
openfang hand activate lead # Daily schedule
openfang hand pause lead # Pause without losing state
openfang hand list # See all available Hands
Build your own with HAND.toml and publish to FangHub.
Architecture: 14 Rust Crates
| Crate | Responsibility |
|---|---|
openfang-kernel | Orchestration, workflows, metering, RBAC, scheduler, budget tracking |
openfang-runtime | Agent loop, 3 LLM drivers, 53 tools, WASM sandbox, MCP, A2A |
openfang-api | 140+ REST/WS/SSE endpoints, OpenAI-compatible API, dashboard |
openfang-channels | 40 messaging adapters with rate limiting, DM/group policies |
openfang-memory | SQLite persistence, vector embeddings, canonical sessions, compaction |
openfang-types | Core types, taint tracking, Ed25519 manifest signing, model catalog |
openfang-skills | 60 bundled skills, SKILL.md parser, FangHub marketplace |
openfang-hands | 7 autonomous Hands, HAND.toml parser, lifecycle management |
openfang-extensions | 25 MCP templates, AES-256-GCM credential vault, OAuth2 PKCE |
openfang-wire | OFP P2P protocol with HMAC-SHA256 mutual authentication |
openfang-cli | CLI with daemon management, TUI dashboard, MCP server mode |
openfang-desktop | Tauri 2.0 native app (system tray, notifications, global shortcuts) |
openfang-migrate | OpenClaw, LangChain, AutoGPT migration engine |
xtask | Build automation |
40 Channel Adapters
Connect agents to every platform:
| Category | Platforms |
|---|---|
| Core | Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Email (IMAP/SMTP) |
| Enterprise | Microsoft Teams, Mattermost, Google Chat, Webex, Feishu/Lark, Zulip |
| Social | LINE, Viber, Facebook Messenger, Mastodon, Bluesky, Reddit, LinkedIn, Twitch |
| Community | IRC, XMPP, Guilded, Revolt, Keybase, Discourse, Gitter |
| Privacy | Threema, Nostr, Mumble, Nextcloud Talk, Rocket.Chat, Ntfy, Gotify |
| Workplace | Pumble, Flock, Twist, DingTalk, Zalo, Webhooks |
Each adapter supports per-channel model overrides, DM/group policies, rate limiting, and output formatting.
27 LLM Providers — 123+ Models
3 native drivers (Anthropic, Gemini, OpenAI-compatible) route to 27 providers: Anthropic, Gemini, OpenAI, Groq, DeepSeek, OpenRouter, Together, Mistral, Fireworks, Cohere, Perplexity, xAI, AI21, Cerebras, SambaNova, HuggingFace, Replicate, Ollama, vLLM, LM Studio, Qwen, MiniMax, Zhipu, Moonshot, Qianfan, Bedrock, and more.
Intelligent routing with task complexity scoring, automatic fallback, cost tracking, and per-model pricing.
16 Security Systems
Defense in depth — every layer independently testable:
- Zeroizing memory for secrets
- Environment clearing (
env_clear()) - Path traversal prevention
- Ed25519 manifest signing
- AES-256-GCM credential vault
- OAuth2 PKCE
- RBAC
- Taint tracking
- WASM sandbox isolation
- And 7 more layers
OpenFang vs Alternatives
Category: This tool is an Agent Operating System / autonomous agent platform.
| Feature | OpenFang | AutoGPT | CrewAI |
|---|---|---|---|
| Focus | Agent Operating System | Autonomous agent platform | Multi-agent orchestration |
| Stars | 12.5K ⭐ | 182K ⭐ | 45.5K ⭐ |
| License | Apache 2.0 | Other | MIT |
| Language | Rust | Python | Python |
| Binary Size | ~32MB single binary | Requires Python runtime | Requires Python runtime |
| Cold Start | 180ms | ~4 seconds | ~3 seconds |
| Idle Memory | 40MB | ~250MB | ~200MB |
| Autonomous Hands | ✅ 7 bundled, HAND.toml | ✅ Autonomous agents | ❌ Task-based |
| Scheduled Execution | ✅ Built-in scheduler | ❌ | ❌ |
| Channel Adapters | ✅ 40 platforms | ❌ | ❌ |
| LLM Providers | ✅ 27 (123+ models) | ✅ Multi-LLM | ✅ Multi-LLM |
| OpenAI-Compatible API | ✅ 140+ endpoints | ❌ | ❌ |
| Security Layers | ✅ 16 systems | Basic | Basic |
| WASM Sandbox | ✅ | ❌ | ❌ |
| MCP Support | ✅ 25 templates | ❌ | ❌ |
| A2A Protocol | ✅ | ❌ | ❌ |
| Desktop App | ✅ Tauri 2.0 | Web UI | ❌ |
| TUI Dashboard | ✅ | ❌ | ❌ |
| Vector Memory | ✅ SQLite + embeddings | ✅ | ✅ |
| Migration Engine | ✅ OpenClaw/LangChain/AutoGPT | ❌ | ❌ |
| Skills Marketplace | ✅ FangHub (60 bundled) | ❌ | ❌ |
| Test Suite | ✅ 1,767+ tests | ✅ | ✅ |
| Visual Workflow | ❌ | ✅ Builder | ❌ |
| Role-Playing Agents | ❌ | ❌ | ✅ Core design |
When to choose OpenFang: You want an Agent Operating System — not a framework, but a full runtime where autonomous Hands work for you 24/7, on schedules, across 40 channels, with 16 security layers, a ~32MB binary, 180ms cold start, and 40MB idle memory. Self-hosted, Rust-powered, production-grade.
When to choose AutoGPT: You want the most popular autonomous agent platform with a visual workflow builder, large community, and extensive ecosystem. Python-based, heavier resource usage, but massively adopted.
When to choose CrewAI: You want role-playing multi-agent orchestration where specialized agents collaborate on tasks. Python-based, MIT license, focused on collaborative intelligence rather than autonomous operation.
Quick Start
# Install (macOS/Linux)
curl -fsSL https://openfang.sh/install | sh
# Initialize
openfang init
# Start daemon — dashboard at http://localhost:4200
openfang start
# Activate a Hand
openfang hand activate researcher
# Chat with an agent
openfang chat researcher
# Spawn a pre-built agent
openfang agent spawn coder
Conclusion
OpenFang isn't another agent framework — it's an Agent Operating System. Built from scratch in Rust with 137,000+ lines of code, 14 modular crates, 16 security layers, 40 channel adapters, 27 LLM providers, and 7 autonomous Hands that work for you around the clock. The ~32MB binary with 180ms cold start and 40MB idle memory makes Python runtimes look bloated. This is what happens when someone decides agents should be infrastructure, not scripts.