Fix Connection Refused (Errno 111) When Using Ollama with AutoGPT in Docker

Replace localhost with host.docker.internal in the Ollama connection string inside your AutoGPT configuration. When both services run in Docker containers, localhost inside a container refers to the container itself, not the host machine. Use http://host.docker.internal:11434/ to reach Ollama running on the host from the AutoGPT container.

The Full Answer

The error [Errno 111] Connection refused when connecting Ollama to AutoGPT in Docker Compose has a specific cause: Docker container networking. This section explains the root problem and provides the exact fix, along with verification steps.

Why the Error Occurs

When you run Ollama and AutoGPT using Docker Compose, each service runs in its own container. Docker containers have isolated network namespaces. The address localhost (127.0.0.1) inside a container refers to the container's own loopback interface, not the host machine's. So when AutoGPT tries to reach Ollama at http://localhost:11434/, it attempts to connect to port 11434 inside the AutoGPT container, where no service is listening. This produces the [Errno 111] Connection refused error.

As confirmed in the accepted Stack Overflow answer by efremidze, the solution is to use a special DNS name that Docker provides to reach the host from within a container.

The Fix: Use host.docker.internal

Replace localhost with host.docker.internal in the Ollama endpoint URL. The correct URL is:

http://host.docker.internal:11434/

This works because host.docker.internal is a DNS name that Docker automatically resolves to the host machine's IP address from inside any container. It is available on Docker Desktop for Windows, macOS, and Linux (with some configuration).

Step-by-Step Implementation

1. Locate the configuration file in your AutoGPT project where the Ollama endpoint is defined. This is typically in a .env file or a config.py file. Look for a variable like OLLAMA_BASE_URL, LLM_PROVIDER_URL, or OLLAMA_HOST.

2. Change the value from:

   OLLAMA_BASE_URL=http://localhost:11434/
   

   to:

   OLLAMA_BASE_URL=http://host.docker.internal:11434/
   

3. Save the file and restart your Docker Compose services:

   docker-compose down
   docker-compose up -d
   

4. Verify that AutoGPT can now reach Ollama by checking the logs:

   docker-compose logs autogpt
   

   Look for a successful connection message or the absence of the [Errno 111] error.

Verifying Ollama is Running on the Host

Before applying the fix, confirm that Ollama is actually running on the host machine. The original poster in the Stack Overflow question did this correctly:

$ curl http://localhost:11434/
Ollama is running

If you get a different response (e.g., curl: (7) Failed to connect to localhost port 11434: Connection refused), then Ollama is not running. Start it with:

ollama serve

Or ensure the Ollama service is enabled and started on your system.

Alternative: Use the Host's IP Address

If host.docker.internal does not work on your platform (e.g., older Linux Docker installations without the --add-host flag), you can use the host machine's actual IP address. On Linux, you can find this with:

ip addr show | grep inet | grep -v 127.0.0.1

Then set the URL to:

http://192.168.x.x:11434/

Replace 192.168.x.x with your host's IP. This is less portable but works as a fallback.

When to Use Each Solution

• host.docker.internal: Use this as the primary solution. It works out of the box on Docker Desktop (Windows, macOS) and on Linux with Docker 20.10+ (requires docker run --add-host host.docker.internal:host-gateway). It is the cleanest approach.
• Host IP address: Use this if host.docker.internal is not available or you cannot modify the Docker run command to add the host mapping. It is more brittle because the IP may change if the host network configuration changes.

Community-Reported Caveats

• Some users on the referenced GitHub issue (Ollama Issue #703) report that host.docker.internal may not work on Linux if Docker is not configured with the host-gateway option. In that case, you need to add --add-host host.docker.internal:host-gateway to your docker run command or in the docker-compose.yml under extra_hosts:

  services:
    autogpt:
      extra_hosts:
        - "host.docker.internal:host-gateway"
  

• On Windows, if you are using WSL 2, host.docker.internal resolves to the Windows host, not the WSL 2 Linux VM. This can cause confusion if Ollama is running inside WSL 2. In that case, use the WSL 2 VM's IP address (found via ip addr show inside WSL 2).

• The error [Errno 111] Connection refused can also appear if Ollama is bound to a different port or interface. By default, Ollama listens on 127.0.0.1:11434. If you changed the port or bound it to a different interface (e.g., 0.0.0.0), adjust the URL accordingly.