Use OpenClaw With OLLM

This guide explains how to configure OpenClaw to use OLLM as its model provider.

By configuring OLLM as your backend, all OpenClaw agent interactions will be routed through OLLM’s OpenAI-compatible API endpoint.

What is OpenClaw?

OpenClaw is an open-source AI agent platform that enables conversational AI across messaging platforms such as Telegram, Discord, Slack, Signal, iMessage, and WhatsApp.

It supports multiple LLM providers through a provider configuration system. OLLM can be configured as a custom OpenAI-compatible provider.

Setup

Because OLLM exposes an OpenAI-compatible API, it must be added manually as a provider inside your openclaw.json configuration.

Step 1: Get Your OLLM API Key

1. Log in to your OLLM dashboard.

2. Navigate to the API Keys section.

3. Generate a new API key.

4. Copy the key securely.

Step 2: Configure OLLM in openclaw.json

Edit your ~/.openclaw/openclaw.json file and add OLLM as a provider.

{
 "env": {
   "OLLM_API_KEY": "your-api-key"
 },
 "models": {
   "mode": "merge",
   "providers": {
     "ollm": {
       "baseUrl": "https://api.ollm.com/v1",
       "apiKey": "${OLLM_API_KEY}",
       "api": "openai-completions",
       "models": [
         { "id": "near/GLM-4.6", "name": "GLM 4.6" },
         { "id": "near/GLM-4.7", "name": "GLM 4.7" }
       ]
     }
   }
 },
 "agents": {
   "defaults": {
     "model": {
       "primary": "ollm/near/GLM-4.6"
     },
     "models": {
       "ollm/near/GLM-4.6": {}
     }
   }
 }
}

Alternatively, set your API key as an environment variable:

export OLLM_API_KEY="your-api-key"

OpenClaw will now recognize the ollm provider and route model requests to https://api.ollm.com/v1.

Step 3: Choose Your Model

Update the primary model under agents.defaults.model.

For example:

"model": {
 "primary": "ollm/near/GLM-4.7"
}

Ensure that the selected model ID matches one available in your OLLM account.

Step 4: Restart OpenClaw

After updating your configuration, restart OpenClaw:

openclaw gateway restart

Your agents will now send requests to OLLM using the selected model.

Model Format

OpenClaw uses the format:

ollm/<model-id>\

Example:

ollm/near/GLM-4.6

Any model exposed by OLLM can be added to the models.providers.ollm.models array and referenced using this format.

Multiple Models with Fallbacks

OpenClaw supports fallback models if the primary model becomes unavailable.

You can configure this as follows:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "ollm/near/GLM-4.6",
        "fallbacks": [
          "ollm/near/GLM-4.7"
        ]
      },
      "models": {
        "ollm/near/GLM-4.6": {},
        "ollm/near/GLM-4.7": {}
      }
    }
  }
}

If the primary model fails, OpenClaw will attempt the fallback in order.

Using Auth Profiles (Recommended)

Instead of storing API keys directly in openclaw.json, you can use OpenClaw’s auth profiles.

Add the following:

{
  "auth": {
    "profiles": {
      "ollm:default": {
        "provider": "ollm",
        "mode": "api_key"
      }
    }
  }
}

Then store your key securely:

openclaw auth set ollm:default --key "$OLLM_API_KEY"

Update your provider config to reference the profile:

"providers": {
  "ollm": {
    "apiKey": "auth:ollm:default"
  }
}

This keeps your API key out of plaintext configuration files.

Monitoring Usage

You can track your OLLM usage in the OLLM dashboard, including:

• Total requests
• Token usage
• Model usage
• Verification status

Common Errors

No API key found for provider 'ollm'

OpenClaw cannot find your OLLM API key.

Fix:

• Verify echo $OLLM_API_KEY
• Confirm ${OLLM_API_KEY} is referenced correctly
• Ensure your auth profile is configured properly

Authentication Errors (401 / 403)

If you encounter authentication failures:

• Confirm your API key is valid
• Verify the base URL is https://api.ollm.com/v1
• Ensure your key has not been revoked

Model Not Found

If a model fails to load:

• Confirm the model ID is correct
• Ensure it is defined under models.providers.ollm.models
• Reference it using ollm/<model-id>

Advanced Configuration

Per-Channel Models

You can configure different models per messaging channel:

{
  "telegram": {
    "agents": {
      "defaults": {
        "model": {
          "primary": "ollm/near/GLM-4.6"
        }
      }
    }
  },
  "discord": {
    "agents": {
      "defaults": {
        "model": {
          "primary": "ollm/near/GLM-4.7"
        }
      }
    }
  }
}

Resources

• OpenClaw Documentation
• OLLM Documentation
• OLLM Dashboard