Connect GPT Image 2 to Hermes Agent and OpenClaw for Fast Image Creation
A practical guide to installing WeChat-Publisher through an agent, preparing the SuperAIAPI key, and turning image generation into a full create-review-improve workflow.
WeChat-Publisher is easier to use than it looks: instead of manually studying every command, you can hand the GitHub repository to Hermes Agent or OpenClaw and let the agent install and check the project for you.
1. What this project is for
It is a WeChat public-account draft publishing tool built for agent workflows, and it can also call GPT-image-2 or nano-banana2 for covers, header images, and article illustrations.
Generate high-quality public-account covers and illustrations with GPT-image-2
Use nano-banana2 for fast image experiments
Push article drafts into the WeChat public-account draft box
Turn Markdown or Feishu documents into article-ready content
https://github.com/aiaiaikkk/WeChat-Publisher
2. Let the agent install it
帮我安装并配置 WeChat-Publisher:https://github.com/aiaiaikkk/WeChat-Publisher
3. What you still need after installation
To use the workflow for real, prepare WECHAT_APPID, WECHAT_APPSECRET, and especially SUPERAIAPI_API_KEY.
WECHAT_APPID
WECHAT_APPSECRET
SUPERAIAPI_API_KEY
4. How to get SUPERAIAPI_API_KEY
Log in to superaiapi.com, open the API token page, create a token, and keep the new key. The project already carries the Base URL, so the API key is the main thing you need to pass in.
SUPERAIAPI_API_KEY=your_api_key
5. Ask the agent to write the .env file
Once you have the key, let the agent write it into the .env file and, if you want, run doctor checks afterward.
我已经拿到了 SUPERAIAPI_API_KEY,请帮我写入 WeChat-Publisher 的 .env。
配置完成后,请运行 npm run doctor。
6. Start generating images through HermesAgent or OpenClaw
After the environment is ready, you can describe an image request in plain language. Hermes can rewrite the prompt, call the image model, return the result to chat, and then review the generated image.
7. Recommended workflow
The strongest pattern is not just generate-once. It is write the article, generate the cover, review the result, compare variants, then send the article into a WeChat draft.
Write or refine the article in Feishu
Ask the agent to generate cover images or illustrations
Have the agent convert the content into Markdown or HTML and create the draft
Review manually in WeChat before final publication