How to configure Ollama for concurrent multi-user access

What this does

By default Ollama runs on localhost. This guide configures it as a network service with concurrent request handling, queue management, and user isolation for team use.

Steps

1. Bind Ollama to all network interfaces.

   # Linux/macOS
   OLLAMA_HOST=0.0.0.0:11434 ollama serve
   

   On Windows:

   $env:OLLAMA_HOST="0.0.0.0:11434"; ollama serve
   

2. Persist the binding as a system service.

   sudo systemctl edit ollama
   

   Add:

   [Service]
   Environment="OLLAMA_HOST=0.0.0.0:11434"
   

   Then:

   sudo systemctl daemon-reload && sudo systemctl restart ollama
   

3. Set concurrency limits to prevent resource exhaustion.

   sudo systemctl edit ollama
   

   Add:

   Environment="OLLAMA_MAX_CONCURRENT_REQUESTS=8"
   Environment="OLLAMA_MAX_QUEUE=16"
   Environment="OLLAMA_NUM_PARALLEL=4"
   

4. Add authentication via reverse proxy. Install nginx and create a .htpasswd file:

   sudo apt install nginx apache2-utils
   sudo htpasswd -c /etc/nginx/.htpasswd user1
   

   Configure nginx:

   server {
       listen 11435 ssl;
       location / {
           auth_basic "Ollama";
           auth_basic_user_file /etc/nginx/.htpasswd;
           proxy_pass http://localhost:11434;
       }
   }
   

5. Test concurrent access from multiple clients.

   # Client 1
   curl -u user1:pass http://server:11435/api/generate -d '{"model":"llama3","prompt":"Hello"}'
   # Client 2
   curl -u user2:pass http://server:11435/api/generate -d '{"model":"llama3","prompt":"Hi"}'
   

Verification

# Send parallel requests
curl -s http://server:11434/api/generate -d '{"model":"llama3","prompt":"test"}' &
curl -s http://server:11434/api/generate -d '{"model":"llama3","prompt":"test"}' &
wait
# Expected: Both requests complete without "server busy" errors

Common failures

• Firewall blocking: Ensure port 11434 is open: sudo ufw allow 11434.
• TLS required: For external access, configure SSL via nginx/Caddy. Never expose unauthenticated Ollama to the internet.
• Queue full errors: Increase OLLAMA_MAX_QUEUE or scale horizontally with a load balancer and multiple Ollama instances.

Operator checkpoint

Before treating this as solved, write down the local runtime, model or package version, hardware/backend if relevant, and the verification output. This keeps the guide useful as a Will-It-Run style decision instead of a one-off command transcript.

Related guides

• How to run multiple models simultaneously on the same system
• How to set up a model switching workflow for different tasks