J
- Category
- AI chat · Self-host
- Cost
- Free
- Country
- community
- Licensing
- FOSS
- Platforms
- macOS · Windows · Linux
Pros and cons
+ what works
- +Native desktop GUI, no Docker or terminal needed to get a chat going
- +Apache 2.0 licensed and developed in the open on GitHub
- +Bundles llama.cpp and pulls models from Hugging Face directly inside the app
- +Can also connect to remote APIs (OpenAI, Anthropic, Gemini, Groq, Mistral) when local hardware falls short
- +Telemetry is opt-in and chat content is excluded from collection by policy
- +Available via Microsoft Store and Flathub in addition to direct downloads
− watch out for
- −License changed from AGPL-3 to Apache 2.0 in 2024, so older third-party copies may carry the stricter copyleft
- −Local model quality depends on hardware; 16 GB RAM is a realistic floor and larger models want a GPU
- −Smaller plugin and integration ecosystem than Open WebUI
- −Pointing Jan at a remote API hands prompts to that vendor; the local-only guarantee only holds for local models
- −Multi-user, RBAC, and server features are not the focus; this is a single-user desktop app
Privacy notes
Conversations, model files, and settings stay on the local machine. Chat history, chat settings, and model choices are never collected, regardless of analytics opt-in. Basic app usage telemetry is opt-in on first launch and routed through PostHog EU under a random ID; it can be toggled off in settings. If a user configures a remote API (OpenAI, Anthropic, Groq, etc.), prompts naturally travel to that provider under their terms. Jan is built by Menlo Research in Singapore; since the app is local, user data stays under the operator's own jurisdiction.
Tags
#foss · #apache-2 · #local-first · #desktop · #llama-cpp · #openai-compat · #gui
Comments (0)
No comments yet. Be the first.