🤖 LibrAgent

A lightweight, stateful platform for autonomous AI agents.

한국어 | 简体中文 | 日本語 | Français | Español | Deutsch | Português

LibrAgent is a local-first agent runner designed to maintain context across interactions. Unlike stateless clients, it keeps browser tabs and terminal sessions alive between turns, allowing agents to work more fluidly within a persistent workspace.

It implements open standards like MCP (Model Context Protocol) and Skills to remain modular and extensible.

---

Why LibrAgent?

The goal of this project is to make autonomous agents accessible. Many existing tools remain trapped behind terminal commands and manual JSON configurations, creating a gap that excludes many potential users. LibrAgent aims to bridge this gap by providing a local-first environment where anyone can deploy and manage agents without needing to be a developer.

---

🎬 Demo

Browser automation and shell execution in a single, stateful workflow.

---

Core Features

1. Persistent Workspace

Agents operate within a long-lived environment rather than spawning fresh processes for every turn.

• Live Webview: Real-time browser automation using Tauri webviews. Sessions and cookies persist across turns.
• Unified Terminal: A persistent, sandboxed shell (Python/Node.js supported) that shares state with the workspace.

2. Multi-Agent Orchestration

LibrAgent allows agents to delegate tasks to specialized sub-agents.

• Assistants: Manage agent profiles with unique system prompts and tool configurations.
• Swarm Intelligence: Parent agents can spawn, message, and await results from sub-agents to solve complex tasks.

3. Extensibility

The platform is designed to be expanded via community standards.

• Extensions (MCP): Full support for the Model Context Protocol. Connect to any MCP server instantly.
• One-Click Presets: Curated catalog for GitHub, Brave Search, etc., available directly in the UI.
• Skills & Playbooks: Reusable behavior snippets and structured workflow templates.

4. Autonomy & Scheduling

• YOLO Mode: Optional autonomous execution for sensitive tools without manual approval.
• Scheduled Tasks: Cron-based automation with workspace-specific targeting and automatic recovery.

5. Context & Metrics

• @mentions: Direct injection of files, skills, or playbooks into chat.
• Multimodal: Handles images and audio for OpenAI, Anthropic, and Gemini models.
• Observability: Real-time TPS metrics and prompt caching hits (for Anthropic/Gemini).

---

📖 Documentation & Guides

LibrAgent is a powerful platform, and we provide detailed documentation to help you get the most out of it.

• Navigation Guide: A map of the application's structure, explaining how to use routes like /assistants (Assistant Profiles) and /playbooks (Workflow Templates) to manage your agents.
• Architecture Guide: Detailed overview of how LibrAgent works under the hood.

---

📦 Installation

Download the latest binaries for Windows, macOS, or Linux from the Release page.

Build from source:

git clone https://github.com/fritzprix/libr-agent
cd libr-agent

For production desktop app builds (Tauri binaries/installers), resolve dependencies and run the Tauri build (which will run the frontend build via pnpm build):

pnpm install
pnpm tauri build

For local development, simply run the Tauri dev server. No need to run pnpm build before pnpm tauri dev — Tauri uses the Vite dev server (beforeDevCommand: pnpm dev), so an extra build just slows you down:

pnpm install
pnpm tauri dev

---

Design Choices

• Local First: Your data and API keys stay on your machine.
• Tauri + Rust: Chosen for security (memory safety), performance, and small binary size.
• SQLite (SeaORM): Used for robust, local persistence of sessions and configurations.

---

Contributing & License

Contributions are welcome. Please see CONTRIBUTING.md.

License: MIT