How-to

Make.com

1. Add a Custom Webhook trigger.
2. Copy its URL into the Agent Webhook URL in your chat's Agent Settings.

3. Add your Make Agent into the canvas.
4. Use {{1.body.Session_Id}} as the conversation key (and {{1.body.Thread_Id}} to segment by thread).

5. Final Webhook Response body — a JSON bundle that echoes the four identifiers plus the Response string from your AI Agent:

   {
     "Org_Id":     "{{1.body.Org_Id}}",
     "Chat_Id":    "{{1.body.Chat_Id}}",
     "Session_Id": "{{1.body.Session_Id}}",
     "Thread_Id":  "{{1.body.Thread_Id}}",
     "Response":   "{{<ai-node>.output}}"
   }

6. The response body must be JSON.
7. Make's default response is the plain text Accepted, which the widget cannot parse.
8. In your Webhook Response module, set the body exactly to the JSON template shown in step 5 above.
9. Set Advanced → Custom headers → Content-Type: application/json.
10. Important: stringify your Agent's response before putting it in Response.

n8n

1. Add a Webhook node (HTTP Method: POST).
2. Open the Production URL tab and copy that URL into the Agent Webhook URL in your chat's Agent Settings.
3. That URL has to be reachable from the internet. Chat4Agent calls your webhook from its own servers, not from your browser.
   On n8n Cloud it already is, and it looks like https://<your-instance>.app.n8n.cloud/webhook/<identifier> — for example https://acme.app.n8n.cloud/webhook/abc123-def456-ghi789. Copy it as it is.
   On a self-hosted n8n it usually is not: http://localhost:5678/… means "this machine", and a private address such as 192.168.x.x or 10.x.x.x only exists inside your network — we cannot call either. Publish n8n under a public hostname first, over HTTPS, and paste that address instead. If you do not manage the network yourself, this is the part to hand to whoever does: they need to point a public name at the machine and forward the port through to n8n.

4. Add an IF node right after. Condition: {{ $json.headers["x-chat4agent-secret"] }} equals the Secret from Agent Settings.

5. Wire the false branch to a Respond to Webhook returning status 401.
6. Wire the true branch to your AI node, using {{ $json.body.Session_Id }} as the conversation key (add {{ $json.body.Thread_Id }} if needed).

7. Final Respond to Webhook body — echo the four identifiers and return the AI output in Response:

   {
     "Org_Id":     "={{ $('Webhook').item.json.body.Org_Id }}",
     "Chat_Id":    "={{ $('Webhook').item.json.body.Chat_Id }}",
     "Session_Id": "={{ $('Webhook').item.json.body.Session_Id }}",
     "Thread_Id":  "={{ $('Webhook').item.json.body.Thread_Id }}",
     "Response":   "={{ $json.output }}"
   }

Step 1Message sent to your Agent

Content-Type: application/json
x-chat4agent-secret: <Secret>
x-make-secret:       <Secret>    # same value — alias for Make
x-make-apikey:       <Secret>    # same value — alias for Make (API key variant)
x-n8n-secret:        <Secret>    # same value — alias for n8n

{
  "Org_Id":     "<organization id that owns this chat>",
  "Chat_Id":    "<chat id being served>",
  "Session_Id": "<unique per visitor, rotates every 1h>",
  "Thread_Id":  "<conversation thread id>",
  "Message":    "<visitor text, Markdown allowed>"
}

Step 2Response sent to Chat4Agent

Content-Type: application/json

{
  "Org_Id":     "<same as request>",
  "Chat_Id":    "<same as request>",
  "Session_Id": "<same as request>",
  "Thread_Id":  "<same as request>",
  "Response":   "<agent reply text, Markdown allowed>"
}

5. The Response field must be a JSON string (stringified).
6. Markdown syntax (headings, bold, italics, lists, links, code blocks) is preserved through the widget.
7. Echo the four identifiers back: it keeps every reply traceable.