Lucarne provides a "0-Intrusion" mobile control bridge for local AI agents like Codex, Pi, and Claude. It enables users to manage AI sessions remotely via Telegram or WeChat, receiving notifications, approving actions, and resuming work without altering existing projects. This offers seamless, non-disruptive oversight for AI-powered coding and task execution.

How It Works

The system comprises lightweight background daemons (lucarned) and channel-specific adapters for Telegram and WeChat. It monitors local agent processes, pushing progress updates and actionable events (approvals, questions) to the user's chosen mobile platform. Users can then interact via quoted replies (WeChat) or a Telegram console, resuming sessions and managing agents without direct local access. This architecture avoids hooks, skills, or project modifications, prioritizing a zero-impact setup.

Quick Start & Requirements

• Installation: Available via direct script downloads for macOS/Linux (curl -LsSf https://github.com/tuchg/Lucarne/releases/latest/download/lucarned-installer.sh | sh) and Windows PowerShell (powershell -c "irm https://github.com/tuchg/Lucarne/releases/latest/download/lucarned-installer.ps1 | iex"), or recommended via Homebrew (brew tap tuchg/Lucarne https://github.com/tuchg/Lucarne && brew install lucarned).
• Initialization: Run lucarned init to configure agent selection (Claude, Codex, Gemini, Copilot, Pi), Telegram Bot Token, and WeChat login (QR code). This generates ~/.lucarned/lucarned.yaml.
• Service: Start the background service with lucarned autostart install --start (integrates with macOS LaunchAgent, Windows Task Scheduler, Linux systemd user services). Homebrew users can use brew services start lucarned.
• Prerequisites: Telegram Bot Token, WeChat account, systemd user services on Linux for autostart.
• Docs: Command reference: https://github.com/tuchg/Lucarne/blob/main/docs/commands.md.

Highlighted Details

• Broad Agent Compatibility: Supports reasoning, tool calls, and session management for Claude, Codex, Gemini, Copilot, and Pi.
• Mobile-First Control: Offers a "quote-to-route" mechanism in WeChat and a multi-agent console via Telegram Topics for seamless remote interaction.
• Zero-Intrusion Design: Requires no project changes, hooks, or skills, enabling rapid deployment via QR code scanning.
• Lightweight Operation: Features a low-memory, high-performance resident process with automatic idle agent release.

Maintenance & Community

The README does not detail specific contributors, sponsorships, or community channels (e.g., Discord, Slack). The project includes a roadmap indicating ongoing development.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The permissive MIT license supports commercial use and integration with closed-source applications.

Limitations & Caveats

The roadmap indicates planned support for additional message modes, remote agent environments, and more communication channels (Discord, Slack, etc.). Linux autostart relies on systemd user services; manual execution is required for non-systemd systems. Telegram integration requires specific chat configurations (Topics/thread mode or bot's own topic mode).