RTMS participant-initiated start: pending requests, missing rtms_started, requester popup

Hi all,

Our General App starts RTMS on behalf of a participant: PATCH /v2/live_meetings/{meetingId}/rtms_app/status with action: "start" and participant_user_id. The host approves, and we wait for meeting.rtms_started. If it doesn’t arrive within 2 minutes, we send action: "stop".

Questions:

  1. Withdrawing a pending request. Sending stop before the host approves returns 13277 "...because it has not been started". The host prompt stays up, and a late approval still starts RTMS. How do we cancel a pending request?

  2. Prompt expiry. How long does the host approval prompt stay valid? Is this configurable?

  3. No prompt after an interruption. Our server restarted mid-stream, and we received rtms_interrupted followed by rtms_stopped 60 seconds later. After that, every new start for the same participant in that meeting returns 2xx, but the host sees no prompt and we get no webhook. A later stop returns 13277. Is this expected? How do we reset that state?

  4. Reconnect window. Can another server instance resume the stream within those 60 seconds using the same rtms_stream_id? Is there a recommended graceful-shutdown pattern?

  5. Approved, but no rtms_started. On a live meeting, the host approved and our stop later succeeded, so Zoom treated the app as started. But meeting.rtms_started never arrived. A new meeting worked fine. When does this happen, and can we detect it?

  6. Requester popup. The participant who triggers the request sees a popup with buttons that do nothing. It stays open after the host approves and shows “0” for permissions. Can we hide it or have it auto-dismiss? Will the “0” be replaced once the app is listed on the Marketplace? Other listed RTMS apps show no requester popup. Is that because they use auto-start?

  7. No rtms.request_status on host Deny. We enabled the rtms.request_status event subscription (scope rtms:read) on the same endpoint that receives meeting.rtms_started. We ran three tests: the participant sends start (2xx), and the host clicks Deny on the prompt. Zoom sent no webhook at all; our endpoint received no request. A later stop returns 13277. Does rtms.request_status (status 1 or 3) fire when the host denies a participant-initiated request? If not, which event should we use to detect the denial? Does adding rtms:read require the app to be re-authorized before Zoom starts sending the event?

Happy to share meeting UUIDs privately. Thanks!

@Samruddhi20 some questions and response inline

  1. Withdrawing a pending request. Sending stop before the host approves returns 13277 "...because it has not been started". The host prompt stays up, and a late approval still starts RTMS. How do we cancel a pending request

There is no cancel request at the moment. If there a need for this feature?

2.Prompt expiry. How long does the host approval prompt stay valid? Is this configurable?

Assume that this will be pending until the host acts on it, or until the meeting ends.

3.No prompt after an interruption. Our server restarted mid-stream, and we received rtms_interrupted followed by rtms_stopped 60 seconds later. After that, every new start for the same participant in that meeting returns 2xx, but the host sees no prompt and we get no webhook. A later stop returns 13277. Is this expected? How do we reset that state?

Let me check on this limbo state. In a typical scenario, there will be RTMS interrupted webhook event. Use this event to flag the specific streamID in a list of monitored streams that this specific stream should be reconnected by any available service within the next 60 seconds.

4.Reconnect window. Can another server instance resume the stream within those 60 seconds using the same rtms_stream_id? Is there a recommended graceful-shutdown pattern?

Yes another server instance can resume the stream.

5.Approved, but no rtms_started. On a live meeting, the host approved and our stop later succeeded, so Zoom treated the app as started. But meeting.rtms_started never arrived. A new meeting worked fine. When does this happen, and can we detect it?

Could you elaborate this a little? I’m confused about the “stop” succeeding with the combination of meeting.rtms_started. Usually if meeting.rtms_started doesn’t arrive, you might need to reauthorize and add the app again in on marketplace.zoom.us. There is also a events dashboard in marketplace.zoom.us which you can use to troubleshoot API calls and Events success/failures.

6.Requester popup. The participant who triggers the request sees a popup with buttons that do nothing. It stays open after the host approves and shows “0” for permissions. Can we hide it or have it auto-dismiss? Will the “0” be replaced once the app is listed on the Marketplace? Other listed RTMS apps show no requester popup. Is that because they use auto-start?

I need some screenshot for this. Could you also share what OS and type of client the participant is using? (Web ? Zoom Client on Windows? )

7.No rtms.request_status on host Deny. We enabled the rtms.request_status event subscription (scope rtms:read) on the same endpoint that receives meeting.rtms_started. We ran three tests: the participant sends start (2xx), and the host clicks Deny on the prompt. Zoom sent no webhook at all; our endpoint received no request. A later stop returns 13277. Does rtms.request_status (status 1 or 3) fire when the host denies a participant-initiated request? If not, which event should we use to detect the denial? Does adding rtms:read require the app to be re-authorized before Zoom starts sending the event?

Yes, the rtms.request_statusis intended for this purpose. The participant using the API must have the scope for this request_status event, the participant must also be invited to the meeting, and the rtms.request_status is only triggered when RTMS start is triggered via API.
Any additional scopes and changes requires reauthorization of the app.

Thanks for the detailed answers! Follow-ups inline:

1. Cancel request: Yes, we need this. We start RTMS for a participant via PATCH /v2/live_meetings/{meetingId}/rtms_app/status. If the host doesn’t respond, the requester can’t withdraw the request, and the host can approve much later, after the requester has given up. Today we work around it by calling stop as soon as a late meeting.rtms_started arrives. A cancel action (or an optional expiry on the request) would keep both sides in sync.

2. Prompt expiry: Thanks, understood.

3. No prompt after a stop (new data): This also happens without an interruption:

  1. start → host approves → we receive meeting.rtms_started.
  2. We call action: "stop" (succeeds) → we receive meeting.rtms_stopped with stop_reason: 2. No rtms_interrupted.
  3. About 13 minutes later, a new start for the same participant in the same meeting returns 2xx, but the host sees no prompt and we receive no webhook. stop then returns 13277.

In an earlier meeting, a new start after a stop did prompt the host. The difference we see: in that case the stream had been connected and streaming before the stop, while here we stopped it right after rtms_started, before connecting to the signaling/media servers.

  • Is there per-participant (or per-meeting) state that blocks a new prompt after a stop?
  • Does it depend on whether the stream was ever connected?
  • Is there any way to reset it short of a new meeting?

4. Reconnect: Thanks! A few details:

  • Does the new instance reuse the original server_urls from meeting.rtms_started, or does meeting.rtms_interrupted provide them?
  • Is the window 60 s for signaling and 65 s for media, as in the docs?
  • For a planned shutdown, should the old instance just close its WebSockets (so you send rtms_interrupted), or is there a cleaner handoff signal?

5. Approved but no rtms_started (clarification): Timeline, seen twice on different days:

  1. start → 2xx. The host clicks Approve within about 30 s.
  2. No meeting.rtms_started arrives. Our endpoint was healthy and received other events.
  3. About 2 minutes after the request we call action: "stop" and it succeeds. Before approval, stop returns 13277, so we read the success as Zoom considering the stream started.

New observation: in these cases the requester-side popup also didn’t appear, while it does appear when the flow works. Does that point to the request not being fully registered on the participant’s client? We’ll check the Events dashboard for those windows and re-authorize the app. What should we look for there?

6. Requester popup: Screenshot attached.

  • Requester: [Chrome OS], [Zoom web]

  • Host: [Chrome OS], [Zoom web]

    Unable to add images/links directly in the chat.

Two related UI observations:

  • After our API stop (and meeting.rtms_stopped delivered), the host’s desktop client still showed the notetaker/app as enabled. Is that expected?
  • In one test, the host’s approval prompt disappeared by itself after about 2 minutes without the host acting. Is there a client-side timeout?

7. rtms.request_status: Thanks, we’ll re-authorize and retest.

Additional questions

  • stop_reason codes: where can we find the full list of stop_reason values in meeting.rtms_stopped? We see 2 after our API stop.
  • Status lookup: is there an API to read the current RTMS state for a participant in a meeting (pending / approved / started / stopped)? That would let us detect the stuck states in 3 and 5 instead of waiting.

Happy to share meeting UUIDs, timestamps, and rtms_stream_ids by DM. Thanks again!