OpenClaw Skills | The Largest Library of OpenClaw Extensions

Ollama is a local LLM runtime that allows you to run open-source models more conveniently on your own machine. OpenClaw integrates with Ollama's OpenAI-compatible API; when you enable it via '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'OLLAMA_API_KEY'</code>' (or an auth profile) and do ''not'' explicitly configure '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'models.providers.ollama'</code>', OpenClaw can ''automatically discover models that support tools''.

Tutorial.step

Quick Start

1. Install Ollama: https://ollama.ai

2. Pull a model:

Bash

ollama pull llama3.3

ollama pull qwen2.5-coder:32b

ollama pull deepseek-r1:32b

3. Enable Ollama for OpenClaw (any value works; Ollama doesn't need a real key):

Bash

export OLLAMA_API_KEY="ollama-local"


openclaw config set models.providers.ollama.apiKey "ollama-local"

4. Use Ollama models:

Json5

{
  agents: {
    defaults: {
      model: { primary: "ollama/llama3.3" },
    },
  },
}

Tutorial.step

Model Discovery (Implicit Provider)

When you set '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'OLLAMA_API_KEY'</code>' (or an auth profile) and ''do not'' define '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'models.providers.ollama'</code>', OpenClaw automatically discovers models from your local Ollama instance at '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'http://127.0.0.1:11434'</code>':

- - Query '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'/api/tags'</code>' and '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'/api/show'</code>'

- - Only keep models that report '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'tools'</code>' capability

- - Mark models as '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'reasoning'</code>' when they report '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'thinking'</code>'

- - Read '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'contextWindow'</code>' from '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'model_info["<arch>.context_length"]'</code>' when possible

- - Set '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'maxTokens'</code>' to 10x the context window

- - Set all costs to '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'0'</code>'

This avoids manual model entry maintenance while keeping the catalog consistent with Ollama's capabilities.

View available models:

Bash

ollama list
openclaw models list

When adding new models, just pull with Ollama:

Bash

ollama pull mistral

New models will be automatically discovered and available for use.

If you explicitly configure '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'models.providers.ollama'</code>', auto-discovery is skipped and you need to manually define models (see below).

Tutorial.step

Configuration

Tutorial.step

Basic Configuration (Implicit Discovery)

The simplest way to enable is to set an environment variable:

Bash

export OLLAMA_API_KEY="ollama-local"

Tutorial.step

Explicit Configuration (Manual Models)

Useful when:

- - Ollama is running on a different host/port

- - You want to force specify context window or model list

- - You want to include models that don't declare tool support

Json5

{
  models: {
    providers: {
      ollama: {
        // OpenAI-compatible API baseUrl needs to include /v1
        baseUrl: "http://ollama-host:11434/v1",
        apiKey: "ollama-local",
        api: "openai-completions",
        models: [
          {
            id: "llama3.3",
            name: "Llama 3.3",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 8192,
            maxTokens: 8192 * 10
          }
        ]
      }
    }
  }
}

If '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'OLLAMA_API_KEY'</code>' is set, you can omit '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'apiKey'</code>' in the provider entry, and OpenClaw will auto-fill it during availability checks.

Tutorial.step

Custom Base URL (Explicit Configuration)

When Ollama is running on a different host or port (explicit configuration disables auto-discovery, so you also need to manually define models):

Json5

{
  models: {
    providers: {
      ollama: {
        apiKey: "ollama-local",
        baseUrl: "http://ollama-host:11434/v1",
      },
    },
  },
}

Tutorial.step

Model Selection

After configuration, all your Ollama models are available:

Json5

{
  agents: {
    defaults: {
      model: {
        primary: "ollama/llama3.3",
        fallback: ["ollama/qwen2.5-coder:32b"],
      },
    },
  },
}

Tutorial.step

Advanced

Tutorial.step

Reasoning Models

When Ollama reports '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'thinking'</code>' in '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'/api/show'</code>', OpenClaw marks the model as having reasoning capabilities:

Bash

ollama pull deepseek-r1:32b

Tutorial.step

Model Costs

Ollama runs locally and is free, so all model costs are set to $0.

Tutorial.step

Context Window

For auto-discovered models, OpenClaw uses the context window reported by Ollama when possible; otherwise defaults to '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'8192'</code>'. You can override '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'contextWindow'</code>' and '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'maxTokens'</code>' in explicit provider configuration.

Tutorial.step

Troubleshooting

Tutorial.step

Ollama Not Detected

Confirm Ollama is running and you set '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'OLLAMA_API_KEY'</code>' (or an auth profile), and ''do not'' explicitly define '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'models.providers.ollama'</code>':

Bash

ollama serve

Also confirm API is accessible:

Bash

curl http://localhost:11434/api/tags

Tutorial.step

No Available Models

OpenClaw only auto-discovers models that report supporting tools. If your model doesn't appear in the list:

- - Pull a model that supports tools, or

- - Manually define the model in '<code className="bg-white/10 px-1.5 py-0.5 rounded text-emerald-300 text-sm">'models.providers.ollama'</code>'

Add new models:

Bash

ollama list  # View installed models
ollama pull llama3.3  # Pull a model

Tutorial.step

Connection Refused

Check if Ollama is running on the correct port:

Bash

ps aux | grep ollama


ollama serve

Tutorial.step

Ollama

Quick Start

Model Discovery (Implicit Provider)

Configuration

Basic Configuration (Implicit Discovery)

Explicit Configuration (Manual Models)

Custom Base URL (Explicit Configuration)

Model Selection

Advanced

Reasoning Models

Model Costs

Context Window

Troubleshooting

Ollama Not Detected

No Available Models

Connection Refused

See Also