How to Test Webhooks in Claude
Use Claude Code or Claude Desktop with OpenWebhook MCP: print an inspect URL, watch, and read the next request in the chat. No account for the first run.
In Claude, webhook testing is an MCP session: the agent prints a public URL, watches it, and shows the next HTTP request in the chat. The tool names live on OpenWebhook MCP tools. This page is only the Claude setup.
The inspector in a browser still works if you do not want MCP: test webhooks online.
Claude Code
npm install --global openwebhook
claude mcp add openwebhook -- openwebhook mcp
A Pro token is optional. Add it only for a permanent slug:
claude mcp add openwebhook \
--env OPENWEBHOOK_TOKEN=owh_live_YOUR_TOKEN \
-- openwebhook mcp
Run /mcp and confirm openwebhook is connected. Ask: “Give me my OpenWebhook inspect URL, then watch it.”
Claude Desktop
Install the Desktop Extension (.mcpb) from a release, or run openwebhook mcp as a stdio server. Leave the token empty for an anonymous URL.
Fire a sample
With watch and wait_for_webhook running, POST to the URL Claude printed, or paste that URL into Zapier, Stripe, or GitHub.
curl -X POST \
'https://hooks.openwebhook.co/YOUR-UUID/claude' \
-H 'content-type: application/json' \
--data '{"from":"claude"}'
Claude should return method, path, headers, and body. If /mcp is disconnected, the tools never run.
When to use the CLI
Use listen when a local HTTP server must receive the request. Use MCP when you want the payload inside Claude to write a handler.
