azure-proxy

What This Skill Does

The azure-proxy skill for OpenClaw is a specialized middleware designed to bridge the compatibility gap between the OpenClaw agent architecture and Azure OpenAI's specific API requirements. OpenClaw, by default, constructs API requests in a manner that conflicts with the URL structure required by Azure OpenAI, which mandates an 'api-version' query parameter appended to the deployment path. This lightweight Node.js proxy acts as an intermediary, intercepting standard OpenAI-formatted requests from OpenClaw and reformatting them into the valid structure required by your Azure deployment. By running this service locally, you enable seamless integration with Azure OpenAI, allowing you to bypass 404 errors and fully utilize your Azure-specific model deployments.

Installation

To begin, ensure you have Node.js installed on your system. Install the skill using the ClawHub command: clawhub install openclaw/skills/skills/benediktschackenberg/azure-proxy. Once installed, configure your environment by exporting your Azure credentials: AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_DEPLOYMENT, and AZURE_OPENAI_API_VERSION. It is recommended to deploy this as a systemd service using the provided template to ensure it remains running in the background. Finally, update your ~/.openclaw/openclaw.json configuration file to point the baseUrl for your model provider to http://127.0.0.1:18790. Ensure you set authHeader: false and explicitly define the api-key in the headers section to match Azure's authentication requirements.

Use Cases

This skill is essential for users who leverage Azure OpenAI credits, such as those provided via Visual Studio subscriptions, to offset AI operational costs. It is particularly valuable for power users who need to delegate subagent tasks to more stable, enterprise-grade Azure endpoints. If you are developing within an enterprise environment that restricts direct access to public OpenAI APIs but provides an Azure OpenAI gateway, this proxy serves as the bridge to maintain agent functionality. It is also the ideal solution for developers experiencing persistent 404 errors when attempting to use Azure deployments directly within standard OpenClaw provider configurations.

Example Prompts

1. "OpenClaw, re-initialize your task queue using the new azure-gpt4o model provider so we can start utilizing our enterprise credits."
2. "Check the status of the local azure-proxy health check endpoint to ensure my model connection is currently active."
3. "Switch the subagent configuration to use the Azure-proxied GPT-4o model for all upcoming data analysis tasks."

Tips & Limitations

When using this skill, always verify your deployment name in the Azure portal as it must match your AZURE_OPENAI_DEPLOYMENT variable exactly. Note that Azure's content moderation policies are often stricter than the standard OpenAI defaults; you may encounter 'Content Filter' errors for prompts that would otherwise be accepted by standard models. If you see a 401 Unauthorized error, verify that your API key is correctly scoped and that your resource name in the endpoint string is accurate. This tool is designed for internal, local networking; ensure your firewall settings allow traffic on port 18790 if you intend to run the proxy on a different machine than your primary agent.