Free, open-source local MCP gateway: register each server once, share it with every AI client, and cut tool-definition tokens by up to 91%.
Summary
An open-source, offline-first ChatGPT alternative that runs models on your own computer.
Screenshots
Description
Jan is a desktop assistant that runs entirely on your machine by default. It is open source, works offline, and stores conversations as plain files on your disk rather than in someone else's account.
What it does
- Runs local models with a built-in engine, downloading open weights from within the app and using GPU acceleration where available
- Connects to cloud models too — bring an API key for a hosted provider when a local model is not strong enough for a particular task
- Exposes a local API so other tools on your machine can use the model Jan has loaded
- Supports extensions for adding capabilities and integrations
- Keeps your data local: threads, settings, and models live in a folder you control
Why choose it
Jan is a good fit when privacy is non-negotiable, when you work somewhere without reliable internet, or when you simply want a chat client that does not depend on a subscription staying active. Because it is open source and file-based, nothing is locked away if you decide to move on. Free to download for macOS, Windows, and Linux.
Reviews
Similar App Suggestions
Agentic IDE that runs coding agents, self-healing QA tests and production monitoring in one workspace so fixes reach a PR automatically.
An AI on-call agent that drops read-only probes into running production code, capturing live variable snapshots and stack context without a redeploy or a restart.
A hosted LLM gateway from ngrok: one base URL and one key for public providers, your own keys and self-hosted models, with failover, scoped access and cost analytics.
Cloud-hosted long-horizon AI agents: pick a harness like Claude Code or Codex, pick a model, and reach the same agent from Slack, WhatsApp, iMessage, the web or a CLI.