Summary

lfg provides a private control plane for running and managing AI coding agents like Claude Code, Codex, and OpenCode directly on a user's own VPS or workstation. It targets developers and power users seeking to leverage AI agents within their local codebases, utilizing their own CLIs and credentials securely, offering a significant benefit in privacy and integration.

How It Works

The system transforms a Linux or macOS machine into an agent management hub. It initiates each AI agent within persistent tmux sessions, streaming output to an installable Progressive Web App (PWA) UI. This architecture allows agents to execute directly on the user's machine, granting them access to local repositories, CLIs, and credentials. The core advantage lies in keeping agent execution and data local, enhancing privacy and enabling seamless interaction with the developer's existing environment.

Quick Start & Requirements

• Primary Install: Execute the setup script: curl -fsSL https://raw.githubusercontent.com/BennyKok/lfg/main/scripts/setup.sh | bash (for Ubuntu/Debian VPS or macOS).
• Prerequisites: Bun, tmux, git. At least one supported agent CLI (claude, codex, opencode, hermes) must be available in the system's PATH.
• Optional: Tailscale is recommended for secure, private remote access.
• Links:
  • Website: lfg.apps.omg.dev
  • Local Development: git clone https://github.com/BennyKok/lfg.git && cd lfg && bun install && cp .env.example .env && bun run serve
  • Deployment examples are available in deploy/railway and deploy/hetzner.

Highlighted Details

• Enables AI agents to run directly on your machine, accessing your code, CLIs, and credentials.
• Features a PWA web UI for launching sessions, monitoring output, and interacting with agents from any device.
• Designed for privacy, defaulting to loopback binding and recommending Tailscale for secure remote access.
• Supports optional markdown-defined agents for automated tasks like repo checks and report generation.

Maintenance & Community

The project welcomes issues and pull requests. Specific details regarding maintainers, sponsorships, or dedicated community channels (like Discord/Slack) are not detailed in the provided README.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license permits commercial use and integration into closed-source projects. The project emphasizes optimal performance when run on a machine with local repositories and authenticated agent CLIs, suggesting PaaS deployments are better suited for demos or private network scenarios.

Limitations & Caveats

lfg launches AI agents with shell access on the host machine, necessitating careful security considerations. The control API is intentionally unauthenticated, designed exclusively for loopback and private network access (e.g., via Tailscale); direct exposure to the public internet is strongly discouraged. Some agent CLIs require initial browser or terminal authentication, and API keys must be configured for key-based providers.