Leon is an open-source personal assistant that lives on your own server and carries out tasks through voice or text conversation. You speak or message a request and Leon runs the matching skill module, from checking whether a site is down to downloading a video, optionally without contacting any third party.[1][2]
Quick verdict
Privacy-minded self-hosters and developers who want a voice and text assistant on their own hardware benefit most from Leon. Its modular skill packages, offline Coqui and Flite speech pipeline, and HTTP API cover custom automations without sending audio to a cloud vendor. The constraint is maturity: beta software with paused documentation, English-only offline speech, and no managed hosting or support desk, so non-technical users should choose a maintained commercial assistant instead. If you can run Node.js services and accept beta rough edges, try Leon; if you need turnkey reliability, wait for the official release.[1][2][3][4]
Key features
Two-way voice and text conversation Two-way voice and text conversation: talk to Leon and he talks back, or text him and he texts back, keeping interaction in ordinary language rather than dashboards.[1]
Server-resident core Server-resident core: Leon runs on your own machine and you control your data, including whether any third-party speech service is used at all.[1][2]
Modular packages and skills Modular packages and skills: capabilities ship as modules grouped in packages, such as the checker package with its is-it-down module or the video-downloader package, and anyone can create and share new ones.[1][2]
Fully offline speech pipeline Fully offline speech pipeline: Coqui speech-to-text plus CMU Flite text-to-speech run entirely on your machine, so requests never leave your network when offline mode is configured.[1]
Cloud speech provider choice Cloud speech provider choice: when offline voices are not enough, Leon supports cloud text-to-speech from Google Cloud, AWS, and IBM Watson, and cloud speech-to-text from Google Cloud and IBM Watson.[1]
HTTP query and actions API HTTP query and actions API: external services reach Leon through a POST /api/query endpoint and an actions endpoint secured by an API key, which is how Discord, Signal, Telegram, or Messenger bots talk to your assistant.[1]
Hotword wake listener Hotword wake listener: an independent node process listens for the word Leon, then starts listening for your request, giving hands-free activation without cloud wake-word services.[1]
English and French language content English and French language content: module answers ship with English and French translations and the documentation itself is published in both languages.[1][2]
Who will love it
Privacy-minded server owners who want voice assistance without sending household audio to a cloud vendor, since the whole pipeline can stay on their machine.[1][2]
Developers who enjoy building automations as code, since every new skill is a documented module they can write, test, and share.[1][2]
Tinkerers wiring an assistant into chat platforms or home setups, since the HTTP API and CLI, Docker, or manual installs assume a self-hosted workflow.[1][2]
Who should avoid it
Non-technical users wanting a turnkey assistant, since Leon needs server ownership, command-line or Docker installation, and beta-stage troubleshooting.[1][2]
Teams needing production guarantees or a support desk, since the project is beta, docs updates are paused, and help flows through Discord and GitHub rather than an SLA.[1][2]
Households needing offline speech beyond American English, since the offline recognizer and voice support only en-US today.[1]
Requirements and compatibility
Node.js 16 or newer with npm 8 and Git on the host, since Leon's core runs on the Node.js runtime and its dependencies install through npm.[1]
A generated HTTP API key stored in the environment file for outside access, with the option to disable HTTP exposure entirely when only local interaction is wanted.[1]
About 1 GB of disk for the offline speech model plus provider credentials only when cloud speech is chosen, so offline and cloud setups have different provisioning needs.[1]
Questions worth asking
What is the best Leon alternative for non-technical users?
A maintained commercial assistant or the assistant built into your phone is the safer choice: Leon needs its own server, command-line or Docker installation, and hands-on configuration, and it is still beta software. That setup assumes comfort with servers that non-technical users should not have to acquire.[1][2]
Is Leon really no-code?
How much does Leon cost per month?
Sources and editorial review
Source research. Reviewed on October 11, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- Leon homepage
official · Observed October 11, 2026
Supports identity, license, server-resident positioning, modular architecture, AI-concept positioning, and cloud speech provider names.
- Leon docs introduction
official · Observed October 11, 2026
Supports definition, voice/text interaction, offline option, beta state, install prerequisites and paths, Gitpod trial, docs version and language availability.
- Leon docs offline setup
official · Observed October 11, 2026
Supports offline pipeline facts, model size, environment flags, hotword usage, and the en-US limitation.
- Leon docs HTTP API
official · Observed October 11, 2026
Supports HTTP endpoint paths, API-key authentication, third-party integration examples, and the HTTP disable flag.
- Leon docs packages and modules
official · Observed October 11, 2026
Supports module authoring facts, package naming rules, example packages, and translation-file language evidence.
- leon-ai/leon repository
official · Observed October 11, 2026
Supports canonical repository identity leon-ai/leon as the project's code home.
Compared with these products
No explicit proprietary baseline is recorded.