Remote control for AI coding agents like Codex and OpenCode, Farfield offers a web-based UI for managing, monitoring, and debugging AI sessions from anywhere. It targets developers and researchers needing efficient, secure interaction with AI agents, providing a direct, browser-to-local-backend connection.

How It Works

Farfield comprises a backend server and a frontend web application. Users run the server locally (via npx or bunx), configuring it to interface with specified AI coding agents (Codex, OpenCode, or all). The web UI, accessible at farfield.app, connects directly to this local server, avoiding external intermediaries for enhanced security and privacy. Remote access is achieved by securely exposing the local server's port (default 4311), typically using a VPN solution like Tailscale.

Quick Start & Requirements

• Primary install/run: npx -y @farfield/server@latest or bunx @farfield/server@latest. For development from source, bun install followed by bun run dev or bun run server. Production builds use bun run build and bun run start.
• Prerequisites: Node.js 20+, Bun 1.2+ (for source builds). Requires locally installed Codex or OpenCode agents. Remote access necessitates configuring port 4311 for HTTPS, ideally via a VPN like Tailscale (tailscale serve --https=443 http://127.0.0.1:4311).
• Links: Web UI: https://farfield.app

Highlighted Details

• Thread browser grouped by project.
• Chat view with model and reasoning controls.
• Plan mode toggle for structured interaction.
• Live agent monitoring with interrupt capabilities.
• Debug tab offering full IPC history.
• Supports React Compiler (enabled by default) and production profiling builds.

Maintenance & Community

No specific details on maintainers, community channels (like Discord/Slack), or roadmap were present in the provided README snippet.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license permits broad use, including commercial applications and linking within closed-source projects.

Limitations & Caveats

Secure remote access requires manual VPN configuration (e.g., Tailscale); simpler options are reportedly under development. Development modes like dev:remote expose the server without authentication, mandating use only on trusted networks. Users must have their AI coding agents (Codex/OpenCode) already set up locally.