#Connecting Gateway to external apps
1 messages · Page 1 of 1 (latest)
@vernal echo If you meant the OpenAI “Responses API” shape: OpenClaw does support that, but the endpoint is POST /v1/responses (plural) and it’s disabled by default.
Here’s the quick checklist that usually explains “it disconnects / doesn’t work”:
- Make sure you’re hitting the right endpoint
- âś…
http://<gateway-host>:18789/v1/responses - ❌
/v1/response(will 404)
Docs: https://docs.openclaw.ai/gateway/openresponses-http-api
- Enable it in your OpenClaw config
{
gateway: {
http: {
endpoints: {
responses: { enabled: true },
},
},
},
}
- Send Gateway auth
Authorization: Bearer <gateway token or password>
(Uses your normal gateway.auth.* settings.)
- If you’re streaming (SSE), test with curl first
curl -N http://127.0.0.1:18789/v1/responses \
-H 'Authorization: Bearer YOUR_TOKEN' \
-H 'Content-Type: application/json' \
-d '{"model":"openclaw","stream":true,"input":"hi"}'
If curl streams fine but your app “disconnects”, it’s usually:
- your HTTP client buffering/not handling SSE correctly, or
- a proxy/load balancer timing out long-lived HTTP connections.
- If you actually want Chat Completions compatibility instead
That’sPOST /v1/chat/completions(also disabled by default):
https://docs.openclaw.ai/gateway/openai-http-api
To get you unstuck fast, can you paste: