OpenFang

What is OpenFang?

OpenFang is an open-source Agent Operating System built in Rust, designed to orchestrate autonomous agents that work on schedules, build knowledge graphs, and report to a centralized dashboard. With over 16,800 GitHub stars, it offers a robust platform featuring 7 autonomous Hands, 30 agents, 40 channels, 38 tools, and support for 26 LLM providers. The system is compiled as a single binary, encompassing nine primitives: hands, agents, tools, security, memory, channels, protocols, and a native desktop app.

Key pre-built agents include Clip (video to shorts), Lead (lead generation), Collector (target monitoring), Predictor (forecasting with Brier scores), Researcher (fact-checking with CRAAP), Twitter (X account management), and Browser (web automation). Each agent can be activated, configured, and monitored via the dashboard.

Security is paramount with 16 built-in systems: a WASM dual-metered sandbox, Ed25519 manifest signing, Merkle audit trail, taint tracking, SSRF protection, secret zeroization, HMAC-SHA256 mutual authentication, GCRA rate limiter, subprocess isolation, prompt injection scanner, path traversal prevention, and more. Tool code runs inside WASM with dual metering (fuel + epoch interruption), file operations are workspace-confined, and subprocesses are env-cleared with timeout enforcement. The system also features a 10-phase graceful shutdown.

OpenFang supports 38 native tools plus the Model Context Protocol (MCP) client and server, allowing connection to external MCP servers and exposure of OpenFang tools to other agents. Tools include web search, browser automation, image generation, TTS, Docker, and knowledge graphs. Storage is SQLite-backed with vector embeddings, supporting cross-channel canonical sessions, automatic LLM-based compaction, and JSONL session mirroring. Agents retain context across conversations and channels.

Channels span 40 platforms including Telegram, Discord, Slack, WhatsApp, Teams, IRC, and Matrix, with per-channel model overrides, DM/group policies, rate limiting, and output formatting. The system integrates with 27 LLM providers across four performance tiers (Anthropic, Gemini, Groq, DeepSeek), and agents can be spawned with a single command. Use cases range from orchestration and code review to customer support, making OpenFang a versatile solution for automated workflows.

Key Features

Who is it for?

AI developers, automation engineers, data scientists, DevOps teams, security researchers, content creators, lead generation specialists