OpenClaw GPT Image Proxy

OpenClaw plugin for huwatermelon GPT image proxy. It exposes a gpt_image_generate tool that supports both text-to-image and image-to-image through the proxy async job API.

It uses the proxy async job API:

• POST /v1/images/jobs
• GET /v1/images/jobs/{job_id}

This avoids Cloudflare's long synchronous request timeout.

Install

From npm:

npm install -g openclaw-gpt-image-proxy
openclaw plugins install openclaw-gpt-image-proxy

From a local checkout:

openclaw plugins install /path/to/openclaw-gpt-image-proxy

From a tarball:

npm pack
openclaw plugins install ./openclaw-gpt-image-proxy-0.1.0.tgz

Restart the OpenClaw gateway after installing.

Configure

In ~/.openclaw/openclaw.json:

{
  "plugins": {
    "entries": {
          "openclaw-gpt-image-proxy": {
        "enabled": true,
        "config": {
          "baseUrl": "https://your-proxy.example.com/v1",
          "apiKey": "sk-...",
          "model": "gpt-image-2",
          "timeoutMs": 600000
        }
      }
    }
  }
}

baseUrl is required and accepts either https://host or https://host/v1.

You can also keep the key out of config:

export GPT_IMAGE_PROXY_API_KEY=sk-...
export GPT_IMAGE_PROXY_BASE_URL=https://your-proxy.example.com/v1

Use

Ask OpenClaw to call gpt_image_generate.

Tool parameters:

• prompt: required text prompt
• input_images: optional array of local paths, file:// URLs, HTTPS URLs, or data:image/...;base64,... URLs
• model: optional model override, defaults to gpt-image-2
• output_dir: optional local output directory

Generated images are saved under:

~/.openclaw/generated-images/openclaw-gpt-image-proxy/

The tool returns an OpenClaw image result plus a MEDIA:<path> marker.

If a future OpenClaw runtime exposes registerImageGenerationProvider, this plugin also registers provider id gpt-image-proxy; current OpenClaw releases tested locally only expose the tool API.

Options

• baseUrl: required proxy API base URL
• apiKey: proxy API key
• model: defaults to gpt-image-2
• backend: defaults to cliproxyapi
• pollIntervalMs: defaults to 2000
• maxPollIntervalMs: defaults to 5000
• timeoutMs: defaults to 600000