A cross-device AI automation workstation for everyday users. Uses an adapter-first architecture to integrate Codex, Claude Code, Cline and other local AI tools, executing repetitive digital tasks on real machines with multi-channel access via Web, Feishu, WeChat, Email, and Voice.
RemoteLab is a cross-device AI automation workstation designed for non-technical users, built on the principle of offloading repetitive digital work to AI agents executing on real machines. It adopts an adapter-first executor architecture, defaulting to Codex integration with pluggable support for Claude Code, Cline, and other local AI tools — avoiding lock-in to any single executor.
The session model follows a four-layer structure: Session → Run → App → Share Snapshot. Sessions serve as persistent worker threads with HTTP as the authoritative state path and WebSocket solely for change notifications, ensuring history survives browser disconnections. Apps package validated automation flows into reusable workflows, while Share Snapshots provide immutable read-only session exports.
For access, RemoteLab offers a build-free static Web UI with automatic light/dark theme adaptation, alongside multi-channel connectors for Feishu, WeChat, Email, and Voice — enabling cross-device context continuity across phones, tablets, and desktops. Network exposure supports Cloudflare Tunnel, cpolar, and Tailscale. Interaction features include screenshot/file paste, direct file asset downloads, and AI-powered automatic session naming and grouping.
Typical use cases cover data cleaning, lightweight analysis and report generation, batch file processing, and conditional notifications. The project follows an owner-first single-user model with minimal runtime dependencies (ws, web-push, Lark SDK). Current version is v0.3.1, supporting macOS and Linux only (Node.js ≥ 18).