Leon
Your open-source personal AI assistant with local execution and privacy-first design.
- Source repo
- leon-ai/leon
- Stars
- ★ 18k
- Last updated
- 6d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 43/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Privacy-conscious users who want an AI assistant that works locally without sending data to third parties.
- Not a fit if
- Teams needing a stable, well-documented production system
- Users unable to run Node.js 24 or newer
- Users wanting a turnkey setup with no model or service configuration
- Source review
- 43/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
Leon is an open-source personal AI assistant centered on tools, context, memory, and agentic execution. Originally released in 2019, it is currently being rebuilt for a 2.0 release, transforming into a more capable system. Leon supports three execution modes: smart, controlled, and agent. It can be extended with native skills and agent skills (SKILL.md), and it uses real tools to complete tasks. Leon maintains layered memory to remember preferences and recent context. It supports both local and remote AI providers, balancing privacy and capability. The 2.0 Developer Preview is on the develop branch, while the legacy stable version is on master.
Leon runs a server (server/) that handles routing, memory, context management, HTTP API, and agent/controlled execution. It includes a web app (app/), a UI component library (aurora/), native and agent skills (skills/), bridges (bridges/) connecting Node.js and Python tools, and a TCP server (tcp_server/) for Python services. To use it, run pnpm start and access http://localhost:5366. Leon can execute tasks like searching, productivity, system utilities, media workflows, coding assistance, and voice features. Its architecture follows 'Skills -> Actions -> Tools -> Functions (-> Binaries)', allowing modular extension.
- Privacy-conscious users who want an AI assistant that works locally without sending data to third parties.
- Developers seeking a customizable assistant that can be extended with skills and toolkits.
- Users automating workflows such as search, media processing, or system control, executed by the assistant.
- Users who want the assistant to be context-aware, grounding answers in real machine state or location.
- Contributors interested in joining the 2.0 Developer Preview and shaping the new architecture.
- Users needing both controlled and agent modes for a balance of determinism and adaptability.
How do you install or deploy this agent?
Prerequisites: Node.js >= 24.0.0, and Linux, macOS, or Windows. Recommended to use Volta for Node.js management. Clone: git clone https://github.com/leon-ai/leon.git, then cd leon. Install pnpm: npm install --global pnpm@latest. Install dependencies: pnpm install.
How do you use this agent?
Start Leon with pnpm start. Verify setup with pnpm run check. Access the app at http://localhost:5366. For the 2.0 preview, switch to the develop branch and refer to core/context/LEON.md and core/context/ARCHITECTURE.md for current behavior.
What are this agent's strengths and limitations?
- Supports local AI models, emphasizing privacy and reducing reliance on third-party services.
- Built on 'tools, context, memory, and agentic execution,' more capable than classic intent-based assistants.
- Offers three execution modes (smart, controlled, agent) for flexibility.
- Modular design: skills, toolkits, bridges, and binaries allow easy extension.
- 2.0 is under active development; documentation is incomplete and lags behind the code.
- Requires Node.js >= 24.0.0, which may be an adoption barrier.
- Small contributor base and development largely by the author in spare time may lead to uneven progress.
- If using remote AI providers, there may be cost or privacy trade-offs.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Leon This agent | 43 · Major gaps | CLIFree + model costs | ★ 18k | 6d ago | TypeScript | — |
| MiMoCode | 50 · Major gaps | CLIFreemium | ★ 14k | today | TypeScript | ChatGPT · Codex · Claude Code |
| CORE Personal AI OS | 44 · Major gaps | CLIFree + model costs | ★ 2k | 1mo ago | TypeScript | Codex · Claude Code |
| Compartment | 76 · Good | CLIFree | ★ 583 | 1d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: project emphasizes privacy and local operation, but lacks specific implementation details for least privilege; user confirmation mechanisms exist (e.g., interactive installation), but not all external actions require confirmation; data flow transparency is limited; sensitive data handling is not explicit; many dependencies but no security audit; external effects (e.g., shell execution) exist but not clearly restricted; rollback mechanism not mentioned; source attribution is clear (author Louis Grenard). Deductions: lack of concrete implementation evidence, mostly assertions.
Evidence shows: clear project structure, README consistent with code; many dependencies but no availability guarantee; failure messages not detailed. Deductions: dependency availability unverified, failure handling unclear.
Evidence shows: targeted at developers, supports multiple scenarios; capability boundaries described (e.g., skills, tools); trigger precision not explicit; environment fit documented (multi-OS support). Deductions: trigger precision and scenario details insufficient.
Evidence shows: clear information architecture (directory structure); detailed installation notes; naming stability (version numbers); limited examples and FAQ; known limitations stated (docs not ready); license clear (MIT); versioning exists but changelog not provided; maintenance responsibility clear (author). Deductions: insufficient examples and FAQ, missing changelog.
Evidence shows: output usability not explicit; high marginal value (open source, privacy); cost-benefit not assessed. Deductions: lack of concrete output examples and cost analysis.
Evidence shows: claims traceable (README); limited cross-source verification; facts and inferences not clearly separated. Deductions: lack of independent verification and clear separation.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Project is in 2.0 Developer Preview, documentation incomplete, use with caution.
- Many dependencies without security audit, recommend self-review.
- External operations (e.g., shell execution) not clearly restricted, pay attention to permission control.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
*Leon's TCP server to communicate between the core and third-party nodes.*View on GitHub ↗