Configurations · Level intermediate

Configuration: Ollama local-only, private and with more context

Copyable recipes to turn off Ollama’s cloud features, keep it reachable only from your computer, increase the context and store models on another disk.

By Cognenza editorial team · Updated 15 October 2026 · 3 min read · Sources cited

What it’s for

Ollama runs models on your computer, but it also offers cloud features. These recipes, taken from its official documentation, make sure everything stays on your computer and adapt Ollama to long documents.

1. Turn off cloud features

Create or edit the file ~/.ollama/server.json:

{
  "disable_ollama_cloud": true
}

Or, alternatively, set the environment variable:

OLLAMA_NO_CLOUD=1

2. Keep it on your computer only

By default Ollama listens on 127.0.0.1:11434, that is, only on your computer. Don’t set OLLAMA_HOST=0.0.0.0 unless you know how to protect access from your network.

3. More context for long documents

OLLAMA_CONTEXT_LENGTH=8192 ollama serve

The default context is 4096 tokens. More context means more memory; check with:

ollama ps

4. Store models on another disk

OLLAMA_MODELS=/path/to/another/disk/ollama-models

How to set variables on your system

System	How
Windows	Quit Ollama, search for “environment variables” in Settings, add the variable for your account and reopen Ollama
macOS	`launchctl setenv NAME value` and restart the Ollama app
Linux	Edit the service with `systemctl edit ollama.service`, add `Environment="NAME=value"` and restart

Precautions

After changing anything, restart Ollama so it takes effect.
If an external app uses your Ollama, check its privacy settings too.

Frequently asked questions

What exactly does turning off the cloud do?

According to the official documentation, it disables Ollama’s cloud models and web search, so you only use local models.

Why shouldn’t I use 0.0.0.0?

OLLAMA_HOST=0.0.0.0 makes Ollama accept connections from other computers on your network. Without extra protection, anyone with access could use it. If only you use it on your computer, you don’t need it.

How much context should I set?

The default is 4096 tokens. Raise it only as much as you need: more context requires more memory. With ollama ps you can see the context and whether the model fits on the graphics card.

Keep reading

Sources

Official sources are linked in their original language.

FAQ (documentación oficial) — Ollama
Quickstart (documentación oficial) — Ollama

Change log

15 October 2026: First published.

Educational content, prepared with the help of AI tools, with sources cited. It is not a substitute for professional advice. Tools change fast: if you spot something out of date, write to us at contacto@cognenza.com. How we work.