Keep Ollama Local-Only: Turning Off Cloud Models and Web Search, and What Stays on Your Machine

By Web Asset House Studio · October 29, 2026

Note: Some future links on this site may be affiliate links. We never claim hands-on testing; guides are researched from the public sources listed in each article.

How we wrote this: A summary of Ollama's official FAQ, Cloud and Web search documentation (linked below), checked October 29, 2026. We quote what Ollama says about its own data handling; we haven't audited it. Settings can change between versions, so check the FAQ for your version.

Many people run Ollama precisely so their prompts never leave the computer. Ollama has also added cloud models and a web search API, which do involve ollama.com. If you want a machine that only ever runs local models, for client files, say, Ollama documents a switch for that.

What Ollama says stays local

From the FAQ: "Ollama runs locally. We don't see your prompts or data when you run locally."

Cloud use is different. For cloud-hosted models, the FAQ says Ollama processes your prompts and responses to provide the service but does not store or log that content and never trains on it. It collects basic account information and limited usage metadata that doesn't include prompt or response content. The Cloud page repeats that cloud prompts aren't used to train models and points to Ollama's privacy policy.

So the privacy question isn't "Ollama or not." It's "local model or cloud model."

How cloud features show up

According to the Cloud docs:

In the Ollama app or CLI, cloud models carry a :cloud suffix, for example gemma4:cloud. They don't need to be downloaded, which is one way to tell them apart from local models.
Using them requires signing in to ollama.com (ollama signin), and usage is tracked against a plan.
The web search API requires an ollama.com API key and a free Ollama account; it sends a search query to ollama.com and returns web results, which apps can feed to a model.

If none of that is something you want on a given machine, turn it off.

Turn on local-only mode

The FAQ gives two ways to disable cloud features. Either add this to ~/.ollama/server.json:

{
  "disable_ollama_cloud": true
}

or set an environment variable:

OLLAMA_NO_CLOUD=1

Then restart Ollama. The FAQ says that once it's disabled, Ollama's logs show Ollama cloud disabled: true, which is how you confirm it took effect. Our troubleshooting guide shows where the logs live on each system.

The trade-off, per the FAQ: you lose the ability to use Ollama's cloud models and web search. Downloaded local models keep working.

Setting the variable permanently depends on your system, as the FAQ explains:

Mac app: launchctl setenv OLLAMA_NO_CLOUD "1", then restart the Ollama app.
Linux (systemd): systemctl edit ollama.service, add Environment="OLLAMA_NO_CLOUD=1" under [Service], then systemctl daemon-reload and systemctl restart ollama. Our Linux setup guide walks through systemctl edit.
Windows: quit Ollama, open Settings and search for environment variables, add it to your account's variables, and start Ollama again.

Related settings for a private setup

Keep it on localhost. The FAQ says Ollama binds 127.0.0.1 port 11434 by default, so only the computer itself can reach the API. Changing OLLAMA_HOST to expose it to a network is a separate decision; see our local API guide for the cautions.

Watch the browser origins. The FAQ says Ollama allows cross-origin requests from 127.0.0.1 and 0.0.0.0 by default, and additional origins, including browser extensions, only if you add them to OLLAMA_ORIGINS. If you never added any, web pages and extensions from other origins aren't on the allowed list. Add only extensions you trust.

Don't start at login (optional). The Windows and Mac apps register as a login item. The FAQ says you can turn that off in Task Manager > Startup apps on Windows, or under Login Items > Allow in the Background in macOS Settings, and the setting survives upgrades.

Proxies. If your network requires a proxy to download models, the FAQ says to set HTTPS_PROXY (not HTTP_PROXY, which can interrupt client connections). Downloads still come from Ollama's model library; local-only mode is about cloud models and search, not model downloads.

Local-only isn't the whole picture

Local-only mode controls what Ollama sends. Other apps you connect to it, such as a chat front end, an editor extension or a coding agent, have their own settings and network behavior. Check those too; for example, see Open WebUI and Ollama in VS Code. And the files on the computer are only as private as the computer: disk encryption still matters.

Quick checklist

Use models without a :cloud suffix.
Set disable_ollama_cloud or OLLAMA_NO_CLOUD=1, restart, and look for Ollama cloud disabled: true in the log.
Leave OLLAMA_HOST on its localhost default unless you have a reason.
Review OLLAMA_ORIGINS.
Check the privacy settings of every app connected to Ollama.