Skip to main content
OpenFang Logo

​
The Agent Operating System

OpenFang is an open-source Agent Operating System — not a chatbot framework, not a Python wrapper around an LLM, not a “multi-agent orchestrator.” It is a full operating system for autonomous agents, built from scratch in Rust. Traditional agent frameworks wait for you to type something. OpenFang runs autonomous agents that work for you — on schedules, 24/7, building knowledge graphs, monitoring targets, generating leads, managing your social media, and reporting results to your dashboard. The entire system compiles to a single ~32MB binary. One install, one command, your agents are live.

Quick Start

Get up and running in 5 minutes

Architecture

Understand how OpenFang works

API Reference

Explore the full REST API

Autonomous Hands

Pre-built capability packages

​
Installation

curl -fsSL https://openfang.sh/install | sh
openfang init
openfang start
Dashboard live at http://localhost:4200

​
Key Features

Autonomous Hands

7 pre-built capability packages that run independently on schedules — Clip, Lead, Collector, Predictor, Researcher, Twitter, Browser

16-Layer Security

WASM sandbox, Merkle audit trail, SSRF protection, capability-based access control, and 12 more security systems

40 Channel Adapters

Connect agents to Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Email, and 33 more platforms

27 LLM Providers

123+ models with intelligent routing — Anthropic, OpenAI, Gemini, DeepSeek, Groq, Ollama, and more

MCP & A2A Support

Model Context Protocol client/server and Agent-to-Agent protocol for inter-system communication

Single Binary

~32MB binary with zero dependencies — builds in Rust, runs anywhere

​
Performance Benchmarks

All data from official documentation and public repositories — February 2026.

​
Cold Start Time

  • OpenFang: 180 ms ⭐
  • ZeroClaw: 10 ms
  • LangGraph: 2.5 sec
  • CrewAI: 3.0 sec
  • AutoGen: 4.0 sec
  • OpenClaw: 5.98 sec

​
Idle Memory Usage

  • OpenFang: 40 MB ⭐
  • ZeroClaw: 5 MB
  • LangGraph: 180 MB
  • CrewAI: 200 MB
  • AutoGen: 250 MB
  • OpenClaw: 394 MB

​
Install Size

  • OpenFang: 32 MB ⭐
  • ZeroClaw: 8.8 MB
  • CrewAI: 100 MB
  • LangGraph: 150 MB
  • AutoGen: 200 MB
  • OpenClaw: 500 MB

​
Architecture Overview

OpenFang is organized as a Cargo workspace with 14 crates: 
openfang-cli            CLI interface, daemon auto-detect, MCP server
openfang-desktop        Tauri 2.0 desktop app (WebView + system tray)
openfang-api            REST/WS/SSE API server (Axum), 140+ endpoints
openfang-kernel         Kernel: assembles all subsystems, workflow engine, RBAC
openfang-runtime        Agent loop, 3 LLM drivers, 53 tools, WASM sandbox
openfang-channels       40 channel adapters, bridge, formatter
openfang-wire           OFP peer-to-peer networking with HMAC-SHA256 auth
openfang-memory         SQLite memory substrate, sessions, semantic search
openfang-skills         60 bundled skills, FangHub marketplace
openfang-hands          7 autonomous Hands, HAND.toml parser
openfang-extensions     25 MCP templates, AES-256-GCM credential vault
openfang-migrate        Migration engine (OpenClaw, LangChain, AutoGPT)
openfang-types          Core types, taint tracking, manifest signing
xtask                   Build automation

​
What Makes OpenFang Different?

​
Autonomous Hands

Hands are OpenFang’s core innovation — pre-built autonomous capability packages that run independently, on schedules, without you having to prompt them. 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 reference injected into context at runtime
  • Guardrails — Approval gates for sensitive actions
All compiled into the binary. No downloading, no pip install, no Docker pull.

​
16 Security Systems

OpenFang doesn’t bolt security on after the fact. Every layer is independently testable and operates without a single point of failure:
  1. WASM Dual-Metered Sandbox — Tool code runs in WebAssembly with fuel metering + epoch interruption
  2. Merkle Hash-Chain Audit Trail — Every action is cryptographically linked to the previous one
  3. Information Flow Taint Tracking — Labels propagate through execution
  4. Ed25519 Signed Agent Manifests — Every agent identity is cryptographically signed
  5. SSRF Protection — Blocks private IPs, cloud metadata endpoints, DNS rebinding
  6. Secret Zeroization — API keys auto-wipe from memory
  7. OFP Mutual Authentication — HMAC-SHA256 nonce-based auth
  8. Capability Gates — Role-based access control
  9. Security Headers — CSP, X-Frame-Options, HSTS on every response
  10. Health Endpoint Redaction — Full diagnostics require authentication
  11. Subprocess Sandbox — env_clear() + selective variable passthrough
  12. Prompt Injection Scanner — Detects override attempts in skills
  13. Loop Guard — SHA256-based tool call loop detection
  14. Session Repair — 7-phase message history validation
  15. Path Traversal Prevention — Canonicalization with symlink escape prevention
  16. GCRA Rate Limiter — Cost-aware token bucket rate limiting

​
Community & Support

GitHub

Star the repository

Discord

Join the community

Twitter/X

Follow updates

​
Next Steps

Quick Start

Install OpenFang and spawn your first agent

Core Concepts

Learn how the system works

Autonomous Hands

Explore the 7 built-in Hands

Configuration

Configure models, channels, and security

​
License

OpenFang is open source under the Apache 2.0 and MIT licenses. Use it however you want.
v0.1.0 — First Release (February 2026)OpenFang is feature-complete but this is the first public release. You may encounter instability, rough edges, or breaking changes between minor versions. Pin to a specific commit for production use until v1.0.