OpenFang
The Agent Operating System—an open-source platform built in Rust by RightNow AI. 14 crates, 137K lines of code, ~32MB single binary. OpenFang runs autonomous agents that work for you—on schedules, 24/7—with 7 pre-built Hands (Clip, Lead, Collector, Predictor, Researcher, Twitter, Browser), 40 channel adapters, 38 tools, 27 LLM providers, and 16 security systems.
Key Facts
- Language: Rust
- Footprint: ~32MB binary, ~40MB idle RAM, ~180ms cold start
- Hands: 7 autonomous capability packages—Clip (video shorts), Lead (lead gen), Collector (OSINT), Predictor (forecasting), Researcher (fact-checking), Twitter (X manager), Browser (web automation)
- Channels: 40 adapters—Telegram, Discord, Slack, WhatsApp, Teams, Matrix, IRC, and 33 more
- LLM providers: 27 providers, 123+ models (Anthropic, Gemini, OpenAI, Groq, DeepSeek, etc.)
- Security: 16 systems—WASM dual-metered sandbox, Merkle audit trail, taint tracking, Ed25519 manifests, SSRF protection, secret zeroization, and more
- Memory: SQLite-backed with vector embeddings, cross-channel sessions, LLM-based compaction
- License: MIT
Best For
Teams who want production-grade autonomous agents that run on schedules—lead generation, research, content creation, monitoring—without manual prompting. OpenFang's Hands are pre-built capability packages that work independently. The 16-layer security model and WASM sandbox suit deployments where safety and auditability matter.
Note: v0.1.0 is the first public release (February 2026). Some Hands are more mature than others. Pin to a specific commit for production until v1.0. Includes an OpenClaw migration path.
Compared to OpenClaw
OpenFang shares the same lineage—it offers an OpenClaw migration command. Where OpenClaw is TypeScript and chat-centric, OpenFang is Rust and built around autonomous Hands that run on schedules. OpenFang has 40 channel adapters vs OpenClaw's 13, 16 security layers vs 3, and a smaller footprint (~32MB vs ~500MB install). OpenClaw's ClawHub skill ecosystem remains larger; OpenFang has 60 bundled skills and FangHub.