leon-ai/leon

🧠 Leon is your open-source personal assistant.

What it solves

Leon is an open-source personal AI assistant designed to provide a practical, privacy-aware alternative to cloud-based assistants. It solves the problem of relying on third-party services for personal task automation and information retrieval by allowing users to run the assistant locally with their own context and memory.

How it works

Leon uses a modular architecture based on tools, context, and layered memory. It operates in three distinct modes: smart (automatic selection), controlled (deterministic native skills), and agent (step-by-step planning). The system leverages a hierarchy of Skills, Actions, Tools, and Functions to execute tasks, supporting both local and remote AI providers to balance privacy and capability.

Who it’s for

It is designed for users who want a personal AI assistant that they can host themselves, as well as developers who want to build extensible agentic workflows using native and agent skills.

Highlights

  • Agentic Execution: Supports step-by-step planning and tool use to complete complex tasks.
  • Computer Use: Includes capabilities to operate desktop and browser interfaces and verify visual outcomes.
  • Layered Memory: Maintains durable preferences and recent discussion context to keep interactions grounded.
  • Local-First Approach: Supports local models and local context to ensure user privacy.
  • Modular Extensibility: Uses a system of bridges, toolkits, and binaries to allow for easy expansion of capabilities.

Related

  • Project
    n-WN/SubgroupX

    A multi-user agent platform that allows AI agents to execute tasks on private, non-publicly routable machines via an outbound connector or cloud sandboxes.

  • Project
    beclab/Olares

    Olares is an open‑source personal‑cloud operating system that runs on your own Linux machine (or cluster). Built on Kubernetes, it offers a web UI and a suite of system apps (Files, Vault, Market, Dashboard) plus AI‑specific agents (Lares, OpenClaw). With one‑click model installation, GPU pooling, private networking, and CLI “Agent Skills,” it lets individuals or small teams host LLMs, generative‑AI tools, home‑automation, and collaborative workspaces while keeping data and compute local.

  • Project
    Team-Commonly/commonly

    Commonly is an open‑source, self‑hosted platform that lets humans and AI agents share a persistent workspace (“pod”) with chat, memory, a task board, and tool integrations. Agents join as named seats, keep their own identity and runtime, and can be connected via a daemon, manual CLI, or MCP package. The system includes multi‑LLM routing, GitHub‑synced tasks, scoped grants for external services, and a marketplace of pre‑built agents. It runs on Docker‑Compose locally or via Helm on Kubernetes, is Apache‑2.0 licensed, and provides extensive docs and community support.

  • Project
    tinyhumansai/openhuman

    OpenHuman is an open‑source Rust‑based AI‑agent harness offering a desktop app and a library. It runs hundreds of agents on a cheap VPS by compressing prompts, using a tiny tool‑search model, and keeping the core in a single compiled process. Benchmarks show it’s the fastest, cheapest, and most memory‑efficient among seven comparable harnesses. Features include built‑in memory, voice agents, real‑browser/desktop control, 119 OAuth integrations, 14 messaging channels, and support for local or BYOK models. Install via pre‑built binaries or a simple script; developers embed it with a few lines of Rust code.

  • Project
    Fosowl/agenticSeek

    AgenticSeek is an open‑source, fully‑local AI assistant that can browse the web, write and run code, and plan multi‑step tasks. It runs on your hardware (via Ollama, LM‑Studio, or similar) and stores all data locally, with optional voice interaction and a web UI.