[Question] Persisting / retrieving sub‑agent (node) names in thread history

[austinmw]

Hi,

I am using LangGraph Platform for my backend with a frontend deployed in Vercel.

I stream responses from a langgraph‑supervisor graph and display which sub‑agent is active.
I get the name from debug events in the SSE stream:

// streamMode includes "debug"
for await (const event of stream) {
  if (event.event === "debug" &&
      event.data?.type === "task") {
    const nodeName = event.data.payload?.name;   // e.g. "Email Agent"
    // show "Email Agent: <response>"
  }
}

Debug event shape:

{
  "event": "debug",
  "data": {
    "type": "task",
    "payload": { "name": "<node_name>" }
  }
}

When I later fetch the thread via

GET /threads/{id}
GET /threads/{id}/history

the values.messages (or messages) array contains only standard Agent‑Protocol messages (user, assistant, tool, …).
No node/agent names are present, so after a reload—or on another device—I can’t tell which sub‑agent produced each response.

Questions

1. Is there an existing field or endpoint that already exposes the executing node / sub‑agent name for each step?
2. If not, would it be appropriate to open a feature request for
      • including node‑change data as messages with a custom role/metadata, or
      • adding it to checkpoints / message metadata so it persists in thread history?

Any guidance would be greatly appreciated. Thanks!