Self-host Jan

updated Jul 2026deployed by usprices checked · Jul 2026Install verified on Ubuntu 26.04 · Jul 2026 · how we test

We earn commissions when you shop through the links below. Full disclosure →

Offline-first, open-source ChatGPT alternative that runs LLMs locally on the desktop (llama.cpp based, GPU-accelerated where available), with an optional self-hosted Jan Server for OpenAI-compatible API access across a team.

Key facts

CategoryAI Chat Interfaces

LicenseApache-2.0

StackGo

Min RAM8192 MB

Official imageyes

Difficulty

Sourcegithub.com/janhq/server

Websitejan.ai

Our recommendation

Choose Jan when offline-first is the whole point — it's an Apache-2.0 app that runs LLMs locally with no account and no cloud by default, plus an optional self-hosted Jan Server for OpenAI-compatible API access across a team. It's the privacy-purist pick. If you want a browser-based, multi-user UI rather than a desktop-first app, Open WebUI is the better shape.

Follow the Jan deploy guide →

What you need

Any VPS with at least 8192 MB of RAM
A domain you control — most self-hosted setups need HTTPS in front of them
About an afternoon — budget time for troubleshooting

Where to host itaffiliate disclosure

Hetzner Cloudrun it on

From $23.59/mo · 2 vCPU / 4 GB / 80 GB · EU + US

Get Hetzner Cloud

Kamaterafree trial

From $4/mo · 1 vCPU / 1 GB / 20 GB · US + EU + Asia

Start free on Kamatera →

DigitalOceanalso works on

From $6/mo · 1 vCPU / 1 GB / 25 GB · US + EU + Asia

Deploy on DigitalOcean →

Paid link — we earn a commission if you shop through it.

Install

Run these commands on your server:

# Jan Server — official Docker Compose self-host (janhq/server)
sudo apt-get install -y make golang-go   # the quickstart builds a Go CLI
git clone https://github.com/janhq/server.git jan-server
cd jan-server
cp .env.template .env   # or: make quickstart for the interactive wizard
make up-full  #  or make up-gpu on NVIDIA hardware  →  http://SERVER_IP:8000

deploy guideHow to Deploy Jan on a VPS vps picksBest VPS for self-hostingbenchmarked quarterly · re-priced monthly

What you take on

Jan's offline-first promise is real, but it shapes what the tool is and isn't:

non-negotiableIt's desktop-first. The core Jan experience is a local app (Tauri-based, llama.cpp under it), not a web service — the self-hosted Jan Server exists for team API access, but it's a separate, younger piece to run.

non-negotiableLocal inference wants the hardware. The 8 GB minimum is a starting point; real models want more RAM and are far happier with a GPU — offline doesn't mean lightweight.

non-negotiableIt's a fast-moving project. Jan is still maturing — pin versions and read the release notes, and don't assume feature or config stability across updates.