# Leon (slim)

> Open-source personal assistant by Louis Grenard that the owner runs on their own machine or server. The 2.0 Developer Preview adds an agent loop with tools, layered memory, computer use and a proactive pulse, reached through a web app.

- Full: https://www.anchorterminal.com/tools/leon.md (~7,800 tokens) · this version ~2,030 tokens · JSON https://www.anchorterminal.com/tools/leon.json · canonical https://www.anchorterminal.com/tools/leon
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-12

**E · 41/100 · rank #934 of 986 · #14 in Personal assistants · not agent-ready · confidence medium**


Assessment: Connected-app credentials are encrypted per profile and the Gmail connection is read-only. The 2.0 Developer Preview has no tagged release, runs tools and a 30-minute proactive pulse with no approval step, listens on every network interface with client authentication off, and sends telemetry by default.

## Facts

- Kind: Agent harness · vendor: Louis Grenard (Leon AI) · category: Personal assistants · legal entity: No legal entity found. Copyright Louis Grenard · provenance 68/100
- Packages: npm `@leon-ai/cli`
- Auth: None · pricing: Free · x402: no · licence: MIT
- Probe metrics: not measured yet (probes haven't run)
- Status: 2.0 Developer Preview on the `develop` branch, with no tag or packaged build. `package.json` reads 1.0.0-beta.10+dev. The last tagged core release is 1.0.0-beta.8 of 1 May 2023, on `master`
- Where it runs: On the owner's machine or server, Node.js 24 or later, on Linux, macOS or Windows. State lives under `~/.leon/profiles/<profile>/`. An optional Satellite process on another device runs tools there for a remote server
- Acts without asking: Every tool call in agent mode, which is the default. No approval prompt was found in the source. The pulse runs one self-chosen task each 30 minutes when `pulse_enabled` is true, the default, and later owner declines suppress similar tasks
- Controls: `availability.tools.allowed` and `disabled`, `availability.skills`, `runtime.agent_max_iterations` (256), `runtime.pulse_enabled`. The shell tool refuses a list of patterns such as `mkfs`, `fdisk`, some `dd` forms and a download piped to a shell, and sends commands needing `sudo` to a visible terminal. No sandbox
- Accounts: Gmail (OAuth, `gmail.readonly`, the owner's own Google Cloud client), Notion (personal or integration token, 27 functions including page and block writes and `deleteBlock`), TickTick (token or OAuth, create, update and complete tasks), Spotify (OAuth with PKCE, playback control). No calendar, payment or mail-sending tool was found
- Credentials: Connected-app tokens encrypted with AES-256-GCM in the profile's `connections` folder, with the key in the same profile's `.env`. Model provider API keys sit in that `.env` in plain text. Setup can find Codex, Claude Code, OpenCode, Pi, Hermes Agent, OpenClaw and T3 Code credential files and reuse their keys when the owner chooses
- Memory: Persistent, daily and discussion layers in SQLite under the profile's `memory` folder, mirrored into QMD collections for search, plus `OWNER.md`, a private diary and raw session logs. Facts are extracted from turns automatically
- Deletion: `/session delete` removes a session. The memory tool has `read` and `write` only. The memory manager has `forgetById` and `forgetByQuery` with a seven-day soft-delete window, and no caller for them was found in the server, tools or bridges. Deleting the profile folder removes everything
- Models: llama.cpp and SGLang locally, or OpenRouter, Z.AI, MiniMax, OpenAI, Anthropic, DeepSeek, Moonshot AI, Hugging Face, Cerebras, Groq and Celeris. No default model is set. A ChatGPT subscription or an OpenRouter account can be bound at setup
- Channels: Web app at http://localhost:5366, a Socket.IO interface for custom clients, an HTTP API, and voice with wake word, speech recognition and speech synthesis (all off by default). No messaging-app channel on `develop`. A `feat/messaging-apps` branch exists
- Tools: 48 tools with 228 functions across 16 toolkits in the repository, among them shell, file, browser control, computer use, web search, Gmail, Notion, TickTick and Spotify. Schemas load only when a task needs them. No MCP client
- Telemetry: On unless `telemetry_enabled: false`. Start events carry an instance ID, versions, language, speech providers, operating system, CPU count, memory, Node.js and npm versions. Errors, heartbeats and stops are also sent. The install step registers the instance ID before any setting is read
- Cost: Free. Model usage is billed by the provider the owner picks, or nothing with a local model
- Activity: 17,568 stars, 114 open issues and pull requests, and at least 200 commits between 25 September and 12 October 2026 (GitHub API and the cloned branch, 12 October 2026)

- Scores: Reliability 32, Performance pending, Schema & documentation 33, Agent ergonomics 57, Security & auth 34, Payments & pricing 60, Task success pending, Maintenance & community 15, Transparency & trust 60 · total over the 7 assessed categories
- Why: Reliability, Read with the local-software lines, as software the owner runs. · Schema & documentation, Harness reading of the framework line, as the Goose dossier used. · Agent ergonomics, Harness reading, scored on what a person or pipeline driving it has to supply. · Security & auth, Harness reading of the framework checklist, as the Goose dossier used. · Payments & pricing, Read with the self-hosted rule. · Maintenance & community, The last tagged release of the core is 1.0.0-beta.8 of 1 May 2023. The newest entries on the releases page are pre-release builds of the TCP… · Transparency & trust, MIT (30).
- Sources: 32, open questions: 13, both in the full twin
- Capabilities: assistant.proactive, assistant.self-hosted, assistant.memory, assistant.accounts
- JSON: https://www.anchorterminal.com/api/v1/tools/leon.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/leon.svg` or a link to https://www.anchorterminal.com/tools/leon from a page on getleon.ai or one of its subdomains, or the README of github.com/leon-ai/leon, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Clone the develop branch for 2.0. `npm install --global @leon-ai/cli` and docs.getleon.ai install and describe the legacy 1.0 betas
2. Set `client_interface.auth.enabled: true` and a `LEON_PROFILE_TOKEN` before the machine joins a shared network. The server binds 0.0.0.0
3. Set `telemetry_enabled: false` and `runtime.pulse_enabled: false` in the profile `config.yml` for an unattended or private run
4. Use `availability.tools.allowed` to limit tools. Shell, file writes, computer use and browser control are otherwise available to the agent with no approval prompt
5. Back up `LEON_CONNECTIONS_ENCRYPTION_KEY` from the profile `.env`. Without it the stored connections cannot be read

## Connect

```bash
git clone https://github.com/leon-ai/leon.git && cd leon && npm install --global pnpm@latest && pnpm install && pnpm start
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| ZeroClaw | BB | 70.2 | assistant.proactive, assistant.memory, assistant.self-hosted, assistant.accounts | https://www.anchorterminal.com/tools/zeroclaw.min.md |
| NanoClaw | B | 68.3 | assistant.proactive, assistant.memory, assistant.self-hosted, assistant.accounts | https://www.anchorterminal.com/tools/nanoclaw.min.md |
| Hermes Agent | B | 67.9 | assistant.proactive, assistant.accounts, assistant.memory, assistant.self-hosted | https://www.anchorterminal.com/tools/hermes-agent.min.md |
| OpenClaw | B | 67.1 | assistant.proactive, assistant.accounts, assistant.memory, assistant.self-hosted | https://www.anchorterminal.com/tools/openclaw.min.md |
| IronClaw | B | 66.7 | assistant.proactive, assistant.accounts, assistant.memory, assistant.self-hosted | https://www.anchorterminal.com/tools/ironclaw.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)