OpenFang

What is OpenFang?

OpenFang is an open-source Agent Operating System (Agent OS) written entirely in Rust, designed to build, run, and deploy autonomous agents at a production-grade level. It ships as a single binary containing 30 pre-built agents, 7 autonomous Hands, 38 built-in tools, 40 channel adapters, and support for 26 LLM providers. OpenFang distinguishes itself through a kernel-grade security architecture with 16 layered security systems — including a WASM sandbox, Merkle audit trail, and prompt injection scanner — as well as persistent memory, multi-protocol support (MCP, A2A, OFP), and a native Tauri 2.0 desktop app. It is built for developers and teams who want real autonomous agents that work on schedules and report results, not just chatbots that wait for input.

Key Features

• 7 Autonomous Hands

  Pre-built capability packages — Clip, Lead, Collector, Predictor, Researcher, Twitter, and Browser — that run on schedules, build knowledge graphs, and report progress to your dashboard without requiring constant user input.

• 16-Layer Security Architecture

  Enterprise-grade security including WASM dual-metered sandbox, Ed25519 manifest signing, Merkle audit trail, taint tracking, SSRF protection, prompt injection scanner, and HMAC-SHA256 mutual authentication.

• 40 Channel Adapters

  Deploy agents across Telegram, Discord, Slack, WhatsApp, Teams, IRC, Matrix, and 33 more platforms with per-channel model overrides, DM/group policies, and rate limiting.

• 26 LLM Providers & 30 Pre-Built Agents

  Supports 26 LLM providers including Anthropic, Gemini, Groq, and DeepSeek, with 30 ready-to-deploy agents spanning orchestrators, code reviewers, and customer support roles across four performance tiers.

• MCP, A2A & OFP Protocol Support

  Native Model Context Protocol client and server, Google Agent-to-Agent task delegation, and OpenFang Protocol for peer-to-peer agent networking with mutual authentication.

• Persistent Memory & Workflow Engine

  SQLite-backed vector memory with automatic LLM-based compaction enables agents to retain context across sessions; a multi-agent workflow engine supports fan-out, conditional, and loop execution modes.

Use Cases

• Autonomous Lead Generation : Sales and marketing teams can activate the Lead Hand to autonomously discover, enrich, score, and deduplicate qualified leads on a daily schedule with ICP profiling and CSV/JSON export.
• Content Repurposing : Content creators can use the Clip Hand to automatically transform long-form videos into short clips with captions, thumbnails, and voice-overs, then publish directly to Telegram or WhatsApp.
• Intelligence Monitoring & OSINT : Analysts and researchers can deploy the Collector Hand for continuous OSINT-style monitoring of any target, with change detection, sentiment tracking, and automated knowledge graph construction.
• Multi-Platform Agent Deployment : Developers can deploy a single agent across 40+ messaging and collaboration platforms simultaneously, with platform-specific policies and model routing configured per channel.
• Research & Fact-Checking : Knowledge workers can run the Researcher Hand to autonomously cross-reference sources, fact-check claims using CRAAP evaluation, and generate multi-language cited reports.

FAQs