RightNow-AI's OpenFang Beats LangGraph With a Tiny 32MB Rust Agent OS
A trending Rust-based open-source project called OpenFang packages seven autonomous agents into a single 32MB binary that runs on schedules.
Takeaways
- OpenFang is a trending open-source Rust Agent OS with 17k+ GitHub stars.
- Ships as a single ~32MB binary with 14 crates and 137K lines of Rust code.
- Seven built-in Hands run autonomously on schedules: Clip, Lead, Collector, Predictor, Researcher, Twitter, Browser.
- Enumerates 16 security systems including WASM sandbox, Merkle audit trail, and taint tracking.
- Supports 27 LLM providers, 40 channel adapters, and an OpenAI-compatible API on localhost:4200.
- Project is pre-1.0 with expected breaking changes; see docs.
An open-source project called OpenFang has been climbing GitHub, currently sitting north of 17,000 stars for what its authors call an Agent Operating System. Built by RightNow-AI in Rust, the project pitches itself as an alternative to the usual Python-wrapped LLM orchestrators: a full operating system for autonomous agents written from scratch, compiling to a single ~32MB binary. Whether that framing holds up under scrutiny, the repo has enough substance and traction to be worth a look.
Agents that don't wait for a prompt
The core idea driving the project is what the authors call Hands. Most agent frameworks wait for you to type something. OpenFang instead runs autonomous processes on schedules, 24/7, building knowledge graphs, monitoring targets, generating leads, managing social accounts, and reporting results to a dashboard.
Seven Hands ship built into the binary. Each one bundles a manifest declaring tools and settings, a multi-phase system prompt described as a 500+ word expert procedure, a SKILL.md domain reference injected into context at runtime, and guardrails with approval gates for sensitive actions. In plain terms, a Hand is a packaged agent with its own playbook, tools, and safety rails, activated by one CLI command.