A Rust library for building LLM applications, plus a terminal coding agent built on it whose features are Bevy plugins.
- Stars
- 17,595
- Forks
- 1,478
- Open issues
- 117
- Last commit
- yesterday
What is it?
Leon is a personal AI assistant built around tools, context, memory, and agentic execution. It can run in three modes: smart chooses the approach automatically, controlled follows deterministic native skills and actions, and agent plans step by step. Instead of replying only with text, it calls real tools to get work done.
Leon keeps layered memory covering durable preferences, day-to-day context, and recent conversation, and it can use information about the machine and setup it runs on. It supports both local and remote AI providers, so users can trade off privacy, control, and capability. Its computer use feature can operate desktop and browser interfaces and verify visual outcomes.
The project is in a 2.0 Developer Preview on the develop branch, with a Node.js >= 24 requirement and support for Linux, macOS, and Windows. The older, more stable pre-agentic version lives on the master branch. Built-in skills cover search, productivity, system utilities, media, coding help, and voice features.
Key features
- 01Runs locally by default and can connect to local or remote AI providers.
- 02Offers smart, controlled, and agent execution modes.
- 03Calls real tools and native or SKILL.md-based agent skills to complete tasks.
- 04Keeps layered memory of preferences, daily context, and recent discussion.
- 05Provides computer use for desktop and browser interfaces.
Use cases
- Building a privacy-focused personal assistant that works with local models.
- Automating everyday tasks such as search, productivity, and system utilities.
- Experimenting with agent-style tool use and memory on your own machine.
- Extending an assistant with custom skills and toolkits.
Quick start
- 1
Clone the repository
git clone https://github.com/leon-ai/leon.git - 2
Go to the project root
cd leon - 3
Install pnpm and dependencies
npm install --global pnpm@latest && pnpm install - 4
Run Leon (app at http://localhost:5366)
pnpm start
Things to know
- Requires Node.js 24.0.0 or newer.
- The 2.0 Developer Preview is in progress, the new documentation is not ready, and the docs site may lag behind the code.
Topics
FAQ
- What is leon?
- Leon is an open-source personal AI assistant that runs locally and can use tools, memory and context to complete tasks. Leon is a personal AI assistant built around tools, context, memory, and agentic execution. It can run in three modes: smart chooses the approach automatically, controlled follows deterministic native skills and actions, and agent plans step by step. Instead of replying only with text, it calls real tools to get work done.
- Is leon free to use? What license does it have?
- leon is open source under the MIT license. Check the license terms before using it in commercial products.
- Is leon still maintained?
- Actively maintained. The last commit to leon was on Oct 20, 2026.
- What can I use leon for?
- Building a privacy-focused personal assistant that works with local models. Automating everyday tasks such as search, productivity, and system utilities. Experimenting with agent-style tool use and memory on your own machine. Extending an assistant with custom skills and toolkits.
- How do I install leon?
- Clone the repository: git clone https://github.com/leon-ai/leon.git Go to the project root: cd leon Install pnpm and dependencies: npm install --global pnpm@latest && pnpm install Run Leon (app at http://localhost:5366): pnpm start
- Where can I get leon?
- The source code and installation instructions are on GitHub: https://github.com/leon-ai/leon
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source TypeScript platform for building and deploying durable, long-running AI agents and background workflows with retries, queues and observability.
Open-source desktop AI agent by NetEase Youdao that runs office tasks on your local files and takes commands via IM apps like WeChat and Feishu.
Local-first web search, fetch, crawl and research tools for AI agents over MCP, with no API keys and no per-query cost.
TypeScript, Ruby and Python SDKs for building interactive UI resources that MCP tools can return and hosts can render.
A TypeScript framework for building stateful, long-running LLM agents as graphs, with persistence, memory and human-in-the-loop.