Connect an Ollama server

Who this is for: anyone who runs models on their own Ollama server and wants SpiderGate to route to it.

What this does: explains what SpiderGate needs from an Ollama server, then walks you through exposing it safely and adding its address to your vault.

Before you begin

A machine running Ollama with at least one model pulled. Install it with curl -fsSL https://ollama.com/install.sh | sh on macOS or Linux.
A way to reach that machine from the internet: a public address, a reverse proxy, or a tunnel.
About 15 minutes.

Limits and terms to know first

Warning: SpiderGate calls your Ollama server from SpiderGate's own servers. The form's placeholder, http://localhost:11434, would point SpiderGate at itself, not at your machine. Enter an address that is reachable from the internet.

Warning: Ollama's local API has no login: Ollama says it "does not require authentication". Anyone who can reach the address can use your models and your hardware. Put the server behind a reverse proxy or tunnel you control, and never expose port 11434 openly.

Ollama Cloud keys do not fit here: SpiderGate's Ollama entry takes only a Base URL, with no field for a key.

Steps

Step 1 — Make the server reachable

By default Ollama listens on 127.0.0.1:11434 only. Ollama's FAQ shows how to change that with the OLLAMA_HOST environment variable, and how to put Ollama behind Nginx, ngrok or a Cloudflare tunnel. Choose an option that lets you restrict who can connect.

Step 2 — Add it in SpiderGate

Open Gate → Vault (/dashboard/gate/vault) and click Add Provider. Pick Ollama from Other tools. In Base URL, enter the server's root address, for example https://ollama.example.com, without /v1 or /api. Add a Label and click Add key.

Verify it worked

Ollama has no key, so SpiderGate saves the entry without a check. To test it, give it a slot in one of your own routing aliases naming a model you have pulled, and send one request. The trace shows whether your server answered.

Troubleshoot

Every request fails. SpiderGate cannot reach the Base URL. Open it from a machine outside your network; if it does not answer there, SpiderGate cannot reach it either.

The model is not found. Pull it on the server first, with ollama pull <model>.