02. Start privately
Install Docker Engine and the Compose plugin using Docker’s operating-system instructions. Save your downloaded file in a dedicated directory on the server, then run:
docker compose up -d docker compose ps docker compose logs --tail=100 vane
Alternatively, this local-only quick start uses the official image:
docker run -d --name vane -p 127.0.0.1:3000:3000 -v vane-data:/home/vane/data itzcrazykns1337/vane:latest
Use either Compose or the quick-start command, not both on the same port.
03. Complete model setup
From your own computer, open an SSH tunnel. Replace the hostname and user with your server details. If you changed the port above, change the destination port below.
ssh -L 3000:127.0.0.1:3000 user@your-server
Visit http://localhost:3000 and finish Vane’s setup screen. Choose your provider, models, and credentials there. Never put an API key in the static website or Compose download.
04. Add HTTPS and restrict access
Point your demo subdomain to the VPS. Configure an HTTPS reverse proxy to the loopback port, disable response buffering for streamed answers, and protect the whole instance with an authentication layer before sharing it.
05. Verify before launch
- Make one search and open its citations.
- Restart the container and confirm configuration persists.
- Verify unauthenticated requests cannot reach the instance or its APIs.
- Set model-provider budgets and monitor usage.
- Back up the data volume and test restoring it before upgrading.
Deployment reference: Vane upstream documentation ↗. Reviewed October 21, 2026. Check upstream before upgrading.
Common questions
Is VaneDeploy the official Vane website?
No. This is an independent deployment guide and configuration tool. Vane is maintained by its upstream contributors.
Is self-hosting free?
The software has an MIT license. Your server, storage, backups, and model inference may cost money. Cloud model usage is billed by your provider.
Do all searches stay on my server?
No. Web search reaches external search services. If you select a cloud model, requests also leave your server. Review each provider’s data policy.
Can I run a local model on a small VPS?
It depends on the model and available memory. Start with a cloud provider for a small server, or benchmark a separate Ollama host before buying hardware.