LeAgent

LeAgent is an open-source desktop application that lets you chat with AI models and build automated workflows on your own computer or server. It runs as a self-hosted stack, meaning you install and run it yourself rather than sending your data to a third-party service.

The chat side supports multi-turn conversations with streaming responses. It connects to several AI model providers including DeepSeek, Qwen, OpenAI-compatible APIs, and locally-run models via Ollama. You can swap between providers without reconfiguring anything else. The application also includes a memory system that stores conversational history, semantic facts, and learned procedures so the agent can refer back to earlier context.

On top of chat, LeAgent has a visual workflow editor where you can build automated sequences by connecting nodes on a canvas. Each node represents a tool or action. The built-in toolset covers over 100 operations across categories like web browsing, document processing, database queries, code execution, and generating interface elements. Workflows can be exported as YAML files.

The application supports a skill system where bundles of instructions and tools can be packaged into files, installed from links or archives, and loaded on demand. There is also support for MCP, a protocol for connecting external tool servers.

For installation there are several options. A one-command script handles the full setup for people comfortable with the terminal. A Docker container works for anyone who already runs Docker. Downloadable desktop installers are available for Windows, macOS on both Apple Silicon and Intel, and Linux. The desktop build bundles its own Python runtime so no separate installation is required.

The project also includes an optional sidebar character, described as a desk pet, that displays a customizable animated avatar synchronized with the chat session state. LeAgent is released under the Apache 2.0 license.