Quick Start
Source Files
This page is generated from the following source files:
OpenFang is an open-source Agent Operating System built from scratch in Rust, designed to run autonomous agents that work continuously without manual prompting. Unlike traditional chatbot frameworks or Python-based orchestrators, OpenFang operates as a full operating system for autonomous agents—capable of scheduling tasks, building knowledge graphs, monitoring targets, and delivering results to dashboards around the clock. The entire system compiles to a single ~32MB binary containing 137,000 lines of code across 14 crates, with 1,767+ passing tests and zero clippy warnings (README.md:1-11, README.md:36-49).
System Requirements
OpenFang supports Linux, macOS, and Windows platforms with automatic architecture detection. The installer handles x86_64 (amd64) and aarch64 (arm64) architectures natively.
Supported Platforms:
- Linux: x86_64-unknown-linux-gnu, aarch64-unknown-linux-gnu
- macOS: x86_64-apple-darwin, aarch64-apple-darwin (Apple Silicon)
- Windows: PowerShell installer or MSI package available
Environment Variables (Optional):
OPENFANG_INSTALL_DIR— Custom installation directory (default:~/.openfang/bin)OPENFANG_VERSION— Install a specific version tag (default: latest)
Installation
Unix Systems (Linux, macOS, WSL)
The primary installation method uses a curl-based installer that handles platform detection, download, checksum verification, and PATH configuration automatically:
bash1curl -fsSL https://openfang.sh/install | sh
The installer performs the following operations:
- Detects operating system and CPU architecture
- Fetches the latest release version from GitHub API
- Downloads the platform-specific binary tarball
- Verifies SHA256 checksum if available
- Extracts and installs to
~/.openfang/bin(or custom directory) - Configures PATH in the appropriate shell configuration file
PATH Configuration: The installer automatically detects the user's login shell and updates the corresponding configuration file:
| Shell | Configuration File |
|---|---|
| Bash | ~/.bashrc |
| Zsh | ~/.zshrc |
| Fish | ~/.config/fish/config.fish |
Windows Systems
Windows users should use the PowerShell installer:
powershell1irm https://openfang.sh/install.ps1 | iex
Alternative installation methods for Windows include downloading the MSI installer from GitHub releases or installing via cargo.
Alternative: Install from Source
If pre-built binaries are unavailable for a specific platform, OpenFang can be built from source using Rust's cargo package manager:
bash1cargo install --git https://github.com/RightNow-AI/openfang openfang-cli
macOS-Specific Notes
On macOS, particularly Apple Silicon systems, the installer performs ad-hoc code signing to prevent Gatekeeper from terminating the binary. The installer strips extended attributes (com.apple.quarantine) before signing to avoid "Code Signature Invalid" errors. If automatic signing fails, the installer provides manual instructions for resolving the issue.
Quick Start Path
The fastest way to get OpenFang running requires only three commands:
bash1# 1. Install OpenFang 2curl -fsSL https://openfang.sh/install | sh 3 4# 2. Initialize the system 5openfang init 6 7# 3. Start the dashboard 8openfang start
After executing these commands, the OpenFang dashboard becomes accessible at http://localhost:4200 (README.md:44-49).
Hands Autonomous Agent Capabilities
Hands are OpenFang's core innovation—pre-built autonomous capability packages that run independently on schedules without manual prompting. Unlike traditional agents that wait for user input, Hands work proactively: waking up at scheduled times, performing research, building knowledge graphs, and delivering results to configured channels.
Each Hand bundles four components:
- 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
The 7 Bundled Hands
| Hand | Capabilities |
|---|---|
| Clip | YouTube to vertical shorts pipeline. Downloads videos, identifies best moments, cuts clips with captions and thumbnails, adds optional AI voice-over, publishes to Telegram and WhatsApp. 8-phase pipeline using FFmpeg + yt-dlp + 5 STT backends. |
| Lead | Daily lead generation. Discovers prospects matching ICP, enriches with web research, scores 0-100, deduplicates against existing database, delivers qualified leads in CSV/JSON/Markdown. Builds ICP profiles over time. |
| Collector | OSINT-grade intelligence monitoring. Monitors targets (companies, persons, topics) continuously with change detection, sentiment tracking, knowledge graph construction, and critical alerts. |
| Predictor | Superforecasting engine. Collects signals from multiple sources, builds calibrated reasoning chains, makes predictions with confidence intervals, tracks accuracy using Brier scores. Includes contrarian mode. |
| Researcher | Deep autonomous research. Cross-references sources, evaluates credibility using CRAAP criteria, generates cited reports with APA formatting, supports multiple languages. |
| Autonomous Twitter/X management. Creates content in 7 rotating formats, schedules posts for optimal engagement, responds to mentions, tracks metrics. Includes approval queue. | |
| Browser | Web automation agent. Navigates sites, fills forms, handles multi-step workflows using Playwright bridge. Mandatory purchase approval gate prevents unauthorized spending. |
Hand Management Commands
bash1# Activate a Hand — starts working immediately 2openfang hand activate researcher 3 4# Check progress anytime 5openfang hand status researcher 6 7# Activate lead generation on daily schedule 8openfang hand activate lead 9 10# Pause without losing state 11openfang hand pause lead 12 13# List all available Hands 14openfang hand list
Verification
After installation and initialization, verify the system is working correctly:
1. Binary Verification:
bash1openfang --version
Expected output: Version number (e.g., openfang 0.3.30)
2. Dashboard Access:
Navigate to http://localhost:4200 in a web browser. The dashboard should display the OpenFang interface with agent status and available Hands.
3. Hand Status Check:
bash1openfang hand list
This command lists all available Hands and their current activation status.
(README.md:90-105, README.md:44-49)
Performance Characteristics
OpenFang is optimized for performance across multiple dimensions, significantly outperforming Python-based alternatives:
Cold Start Time (lower is better):
- OpenFang: 180ms
- LangGraph: 2.5 seconds
- CrewAI: 3.0 seconds
- AutoGen: 4.0 seconds
- OpenClaw: 5.98 seconds
Idle Memory Usage (lower is better):
- OpenFang: 40 MB
- LangGraph: 180 MB
- CrewAI: 200 MB
- AutoGen: 250 MB
- OpenClaw: 394 MB
Install Size (lower is better):
- OpenFang: 32 MB
- CrewAI: 100 MB
- LangGraph: 150 MB
- AutoGen: 200 MB
- OpenClaw: 500 MB
(README.md:117-141, README.md:143-163)
Troubleshooting
Issue 1: Binary Not Found After Installation
Symptoms: command not found: openfang after running the installer.
Cause: The shell configuration file was not reloaded, or PATH was not updated correctly.
Solution:
bash1# Reload shell configuration 2source ~/.bashrc # or ~/.zshrc for zsh users 3 4# Verify PATH includes OpenFang 5echo $PATH | grep openfang 6 7# If still not found, manually add to current session 8export PATH="$HOME/.openfang/bin:$PATH"
For Fish shell users:
fish1source ~/.config/fish/config.fish
Issue 2: macOS Gatekeeper Terminating Binary
Symptoms: Binary is killed (SIGKILL) immediately on Apple Silicon Macs.
Cause: macOS Gatekeeper rejects unsigned or improperly signed binaries.
Solution:
bash1# Strip quarantine attributes and re-sign 2xattr -cr ~/.openfang/bin/openfang 3codesign --force --sign - ~/.openfang/bin/openfang
Issue 3: Checksum Verification Failure
Symptoms: Installer reports "Checksum verification FAILED!" during installation.
Cause: Downloaded file is corrupted or tampered with.
Solution:
- Verify network connectivity and retry installation
- If using a proxy or corporate firewall, ensure it's not modifying downloads
- Download manually from GitHub releases and verify checksum independently:
bash
1sha256sum openfang-*.tar.gz
Issue 4: Unsupported Platform or Architecture
Symptoms: Installer reports "Unsupported architecture" or "Unsupported OS."
Cause: Running on an unsupported CPU architecture or operating system.
Solution: Build from source using Rust:
bash1cargo install --git https://github.com/RightNow-AI/openfang openfang-cli
Issue 5: Windows Installation via WSL Fails
Symptoms: Running the Unix installer in WSL triggers a message to use PowerShell instead.
Cause: The installer detects Windows environments (mingw, msys, cygwin) and redirects to the PowerShell installer.
Solution: Use the PowerShell installer from Windows:
powershell1irm https://openfang.sh/install.ps1 | iex
Or install via cargo within WSL:
bash1cargo install --git https://github.com/RightNow-AI/openfang openfang-cli
Next Steps
After completing the quick start:
-
Explore Hands Documentation — Review detailed configuration options for each of the 7 bundled Hands to understand their capabilities and customization options.
-
Configure Channel Adapters — OpenFang supports 40 channel adapters for communication. Configure integrations with Telegram, WhatsApp, Twitter, and other platforms based on deployment requirements.
-
Security Configuration — Review the 16 security systems including WASM sandboxing, Merkle hash-chain audit trails, and capability gates. Configure appropriate security levels for the deployment environment.
-
Custom Hand Development — Create custom Hands by defining
HAND.tomlmanifests with tools, settings, and system prompts. Publish to FangHub for community sharing. -
Production Deployment — For production use, pin to a specific commit until v1.0 release to avoid breaking changes between minor versions. Review the security hardening release notes for the latest stability improvements.