Open inspector →
developer guide

How to Test Webhooks from ChatGPT

Point a Custom GPT Action or an OpenAI webhook at an inspector URL to see the real method, headers, and JSON ChatGPT or OpenAI sends.

ChatGPT does not speak MCP. The usual “test a webhook” job is a Custom GPT Action (your OpenAPI server URL) or an OpenAI platform webhook (account events). Both need a public HTTPS endpoint. Use the inspector first. Inspector basics: test webhooks online.

This is not the MCP tools path. For Claude or Cursor, use those pages.

Custom GPT Action

  1. In GPT Builder → Actions → create an action from an OpenAPI schema.
  2. Set servers[0].url to the inspector, with a path you can recognize:
https://hooks.openwebhook.co/YOUR_UUID/chatgpt
  1. Define a path such as POST /events so the inspector shows /chatgpt/events (or /chatgpt if the schema posts at /).
  2. Authentication: start with None against the inspector. Add API keys only when you move the same schema to a real host.
  3. Use Test on the action, or talk to the GPT so it calls the action.

You should see the method, path, Authorization if you enabled it, and the JSON the model sent. If Test succeeds in the builder but the inspector is empty, the schema servers URL is still a placeholder.

OpenAI webhook endpoint

In the OpenAI dashboard, webhook destinations must be HTTPS. Paste:

https://hooks.openwebhook.co/YOUR_UUID/openai

Send a test event from the dashboard. Confirm the signing headers and body before you point the same destination at production.

When to use the CLI

If the Action should hit a local API, put a listen URL in servers. While the CLI is connected, ChatGPT sees the localhost status.

Related guides