Stack-Cairn/LiveAgent

What it solves

LiveAgent provides a local-first AI agent desktop client that bridges the gap between LLM reasoning and actual system operation. It allows AI agents to move beyond simple chat by giving them the ability to interact with the same local tools and files that a human user would, while maintaining privacy and local control.

How it works

The system consists of a desktop application (built with Tauri, Rust, and React) and an optional Go-based Gateway. The desktop app handles the core logic, including model routing (Claude, Codex, Gemini), local tool execution, and memory management. The optional Gateway acts as a protocol relay, allowing users to control their local agent via a web browser from any location.

Who it’s for

It is designed for developers and power users who want an AI agent capable of managing their local file system, running bash commands, and automating tasks on their own machine without relying entirely on cloud-based agent platforms.

Highlights

• Local Tool Execution: Full file-system access (read, write, edit, delete) and the ability to run non-interactive Bash commands and supervise long-running processes.
• MCP & Skills Ecosystem: Native bridging for the Model Context Protocol (MCP) and a system for loading on-demand "Skills" packages.
• Sub-agent Delegation: Ability to spawn independent sub-agents that work in parallel with isolated worktrees and automatic merging.
• Remote Access: An optional Gateway that enables browser-based control of the local agent via WebSocket and Protobuf.
• Persistent Memory: Integrated Markdown and SQLite FTS for cross-session knowledge management.
• Scheduled Tasks: Support for background cron jobs based on bash, http, or prompts.