Niches · Meeting capture & meeting infrastructure

Stream the meeting live to an agent that acts or speaks during the call

Two or three companies' docs show a full fit.

2companies, full fit
2companies, fit with a limit
47companies carry the tag
13share of the category's work
9first-hand accounts

What counts as done

A transcript or per-participant audio stream arrives over a webhook or WebSocket during the call with each segment tied to a named participant, and the bot can play audio or post chat through a documented endpoint.

Measured, for our own tests: words arrive within 1 second of being spoken, and the stream reconnects after a drop without losing the session. No docs we read state these figures, so they don't count against a listing here.

How agents ask for it: listen to the call and flag risks as they come up, answer questions in the meeting, take live minutes and stop when told, show live speaking time

Does the job 2

The docs show it meets the test above, and nothing we read says otherwise.

ListingWhy
Attendee D
Attendee Labs, Inc.
websocket_settings.per_participant_audio streams audio tagged with participant_uuid, transcript.update webhooks carry speaker_name per utterance, and output_audio, send_chat_message and a bidirectional audio WebSocket let the bot speak and chat. WebSocket reconnects up to 30 times. source
Recall.ai B
Hyperdoc Inc.
Real-time transcription sends utterances with participant name over webhook or WebSocket, separate audio per participant is streamed, and output media, output audio and chat endpoints let the bot act. WebSocket retries 30 times at 3 seconds. source

Does the job, with a limit 2

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

ListingWhy
Meeting BaaS B
SAS SPOKE
Bidirectional WebSocket streaming lets the bot hear and speak, and send-chat-message posts chat, but streamed audio is mixed only, and live transcript segments name the active speaker at that moment, null when unknown; Teams speaker detection lags 1 to 2 seconds. source
MeetStream C
MeetStream Corp.
Live audio arrives as WebSocket frames tagged with speaker_id and speaker_name, live transcripts carry speakerName, and the bot can chat and show images. The docs put live transcripts a second or two after speech, outside the 1-second bar. source

Couldn't check 1

The docs were closed to us, or silent. These carry no mark either way.

ListingWhy
Read AI E
Read AI, Inc.
The plans page names an in-meeting dashboard for the user, but no live API, stream or output endpoint could be checked; the reference returned 403. source

Doesn't do this job 8

Listings in the category that do something else, with the reason.

ListingWhy
Circleback B
Circleback AI, Inc.
API, MCP and webhooks deliver meetings after they end; no live transcript, stream or in-meeting output is documented. source
Fathom C
Fathom Video Inc.
API and webhooks deliver content after a meeting is ready; there's no live transcript, audio stream or in-meeting output. source
Fireflies.ai B
Fireflies.AI Corp.
The Realtime API streams live transcript segments with speaker_name over WebSocket, but no documented call lets the bot speak or post chat, so the success can't be met. source
Granola B
Granola, Inc.
The live transcript shows only inside the app; the API and MCP expose notes after they exist, with no stream or in-meeting output. source
MeetGeek C
AIVISION PRODUCTS SRL
Meetings are readable only after analysis; no live stream, chat or audio output endpoint is in the API or MCP tools. source
Nylas Notetaker API BB
Nylas, Inc.
Notetaker outputs arrive only after the meeting; the docs index lists no live transcript, audio stream or output-audio endpoint, only chat announcements and a static camera image. source
tl;dv D
tldx Solutions GmbH
Meetings and transcripts are readable only once ready; no live stream or in-meeting output exists in the API. source
Zoom Realtime Media Streams (RTMS) C
Zoom Communications, Inc.
Per-participant audio with user_name and diarised transcripts stream over WebSocket, with reconnect guidance, but RTMS is receive-only: no documented endpoint lets the app speak or post chat, so the success can't be met. source

What stops agents finishing

  • live text comes from platform captions 1.5 to 2 seconds late (acc-42)
  • RTMS needs a Developer Pack sold only through sales on some accounts (acc-34)
  • RTMS session stuck connecting with no event (acc-33)
  • speaker ID blank in RTMS audio metadata (acc-31, acc-32)
  • no speech-end or silence event (acc-37)

Around it

Before: meeting-capture.bot-to-transcript. After: voice-agents.tool-call-mid-call, company-knowledge.cited-answer.

Where this comes from

Weight 13 of 100 in Meeting capture & meeting infrastructure: 9 of 56 accounts: four in-meeting agents (acc-22, acc-23, acc-42, acc-43) and five RTMS reports (acc-31, acc-32, acc-33, acc-34, acc-37)

The same as data: /api/v1/use-cases/meeting-capture.live-in-meeting.json, or the MCP tool get_use_case with id meeting-capture.live-in-meeting. Other jobs in this category: /api/v1/niches?category=meeting-capture.

Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person. A mark says what the docs show, not what a run measured.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.