How do I expose Ollama to other machines on my network?

Set OLLAMA_HOST=0.0.0.0:11434 so the server binds every interface, then restart it. Only do this on a trusted LAN, because the API itself has no authentication.

Does Ollama have authentication?

No. The API trusts anyone who can reach the port. For any exposure beyond localhost, put an authenticating reverse proxy in front and restrict network access with a firewall or VPN.

How do I fix CORS errors calling Ollama from a browser?

Set OLLAMA_ORIGINS to the exact web origins allowed to call the API, comma-separated. Avoid the wildcard * in production, which permits any site to reach your server.

Is it safe to put Ollama on the public internet?

Not directly. Because there is no built-in auth, never expose port 11434 to the open web. Use a VPN, or an authenticating proxy with TLS and rate limiting, in front of it.

Build / AI / Ollama

Network Security And CORS

Open Ollama to your network safely with host binding, CORS origins, and an authenticating proxy.

TL;DR

Set OLLAMA_HOST=0.0.0.0 to expose the server on your LAN.
Limit browser callers with the OLLAMA_ORIGINS CORS allowlist.
Never expose port 11434 publicly without an auth proxy.

Jump to section

Bind The Host
CORS With Origins
Protect The Endpoint
Safe Exposure
Tips
Warnings
In Practice
FAQ

Bind The Host

Local Only

The default binds to localhost for safety.

# Default: OLLAMA_HOST=127.0.0.1

Expose On LAN

Bind all interfaces to reach other machines.

export OLLAMA_HOST=0.0.0.0:11434

Pick A Port

Change the port if 11434 is taken.

export OLLAMA_HOST=0.0.0.0:8080

CORS With Origins

OLLAMA_ORIGINS

Allowlist web origins that may call the API.

export OLLAMA_ORIGINS=https://app.me.com

Multiple Origins

Separate several origins with commas.

export OLLAMA_ORIGINS=https://a.com,https://b.com

Avoid Wildcard

Do not use a star in production.

# Avoid OLLAMA_ORIGINS=*

Protect The Endpoint

No Built-in Auth

The API trusts anyone who can reach it.

# Ollama has no API key or login

Auth Proxy

Put an authenticating reverse proxy in front.

# Nginx/Caddy with basic or token auth

Firewall It

Block 11434 from the public internet.

# Allow only LAN or VPN to 11434

Safe Exposure

Use A VPN

Reach the server over a private network.

# Expose via VPN, not the open web

TLS Termination

Terminate HTTPS at the proxy.

# Proxy adds TLS in front of 11434

Rate Limit

Throttle requests at the proxy layer.

# Limit requests per IP at the proxy

Tips

Keep OLLAMA_HOST bound to 127.0.0.1 unless you truly need LAN access, since the API has no authentication of its own.
Set OLLAMA_ORIGINS to your specific trusted domains, so only approved web pages can call the server from a browser.

Warnings

Ollama has no built-in auth; exposing port 11434 to the internet lets anyone run your models and read your prompts.
Setting OLLAMA_ORIGINS=* lets any website call your local server from a browser; scope it to trusted origins instead.

In Practice

Bind the server to the network, restrict CORS origins, and front it with an authenticating proxy.

OLLAMA_HOST=0.0.0.0 makes the server reachable on the LAN.
OLLAMA_ORIGINS limits which browser origins may call it.
A systemd reload and restart apply both variables.
An auth proxy and firewall keep port 11434 off the open web.

# 1. Bind to the LAN (systemd service)
#    Environment="OLLAMA_HOST=0.0.0.0:11434"

# 2. Allowlist trusted browser origins
#    Environment="OLLAMA_ORIGINS=https://app.me.com"

sudo systemctl daemon-reload
sudo systemctl restart ollama

# 3. Never expose 11434 directly; front it
#    with an auth proxy and firewall rules
curl http://localhost:11434/api/tags