Developer Tools for Building AI Agents on Meeting Data
For developers putting an AI agent on live or recorded meetings. Here are the layers you need, how to tell them apart, and what pairs with MeetStream.
Why build an agent on meeting data?
Meeting data is one of the richest inputs an agent can act on. Who said what, when, and in what tone.
That holds across Zoom, Google Meet, and Microsoft Teams. The platforms keep the data inside the call, so a capture layer brings it to your code.
What sits between the call and your agent?
A meeting bot takes a seat in the call and captures the media.
Speech-to-text turns that audio into a transcript you can read.
A voice-agent framework lets your agent listen and talk back.
Summaries, prompts and actions stay yours to write.
MeetStream is the capture and real-time voice layer. One API joins the meeting and streams voice both ways.
The four layers, and why the distinction matters
Almost every tool in this space sits in one of four layers, and they compose.
- Bot capture APIJoins Zoom, Meet and Teams, and returns audio, video and transcripts.
- Speech-to-textTurns audio you supply into text, from a source you bring.
- Voice-agent frameworkRuns the real-time listen and speak loop for your agent.
- Notetaker productA finished app that does the whole job for one end user.
The tools and APIs
A meeting intelligence product often uses all three building layers at once.
| Tool | Category | What it does | Best for |
|---|---|---|---|
| MeetStream | Bot capture API | One API sends a bot into Zoom, Google Meet and Microsoft Teams. It records, transcribes and streams. | Teams who want the bot, the speech engine and the capture in one place. |
| Deepgram | Speech-to-text | A speech recognition API with streaming and pre-recorded models. It converts audio you supply into text. | Low-latency streaming when you already have an audio stream. |
| AssemblyAI | Speech-to-text | A speech-to-text API with streaming and post-call models, plus diarization and entity detection. | High-accuracy post-call transcripts, from audio you already hold. |
| LiveKit | Voice-agent framework | An open-source WebRTC framework. Your agent joins a room, subscribes to microphone tracks, and publishes audio back. | A real-time voice agent inside your own app or a WebRTC room. |
| Pipecat | Voice-agent framework | An open-source Python framework. It runs the voice loop and handles interruptions, and it is transport-agnostic. | Developers who want fine control over each piece of a voice agent. |
| Vapi | Voice-agent framework | A managed platform. You configure the speech, model and voice providers, and it runs the loop for you. | Prototypes and simple voice agents, with less to operate. |
Knowing which layer you are missing is the fastest way to shortlist the right tool.
How to evaluate a tool for your build
Start from what your agent needs to do, rather than from a feature list.
- Reads finished transcriptsA stored transcript is enough, and post-call capture will do.
- Acts during the callYou need real-time capture and low latency to the point of use.
- Talks back in the roomYou need a voice-agent layer sitting on top of the media.
The six things worth checking
- Platform coverage: does it support all of Zoom, Google Meet, and Microsoft Teams
- Real-time or post-call: do you need data during the meeting, or just the final transcript
- Per-participant streams: can you get audio and video split by speaker, not just a mixed track
- Latency: how many milliseconds from spoken word to your server
- Data control: can you bring your own storage, and what are the compliance credentials
- SDK and language support: how quickly can your team integrate it
Build versus buy
The hard part of a meeting bot sits below the model prompt. It is pulling raw audio out of platforms built for people.
Then splitting that audio by speaker in real time and keeping latency low, while the bot holds its seat in the call.
Teams that run their own meeting clients usually spend months on capture, hosting and permissions before any product logic.
Buying a capture API cuts that to an integration you can stand up in hours. Building wins when data sovereignty beats speed.
Real-time versus post-call
Most meeting products want both paths, and a capture layer that does both saves an integration.
- Streaming gives you speedWords within a second or two, which is what live features need.
- Post-call gives you accuracyThe engine has the full audio and can rescore with a language model.
- Use each for its jobA streaming path for in-call features, and a post-call path for the record.
Per-participant audio and video
A single mixed track is enough for a basic transcript. Agents that reason about who said what work better on separate streams.
Separate streams are the hard part to build in house, because conferencing platforms hand out one mixed track.
Check early that your capture API returns a stream per speaker, and per-participant video alongside it.
How MeetStream fits in
MeetStream is the meeting-bot infrastructure layer, with real-time voice built in.
- One request, one botSend a meeting link and a bot name, and a bot joins the call.
- Media and text come backRecording, speaker-identified transcripts, and per-participant audio and video.
- Real-time audio outOver WebSocket at roughly 200ms, as binary PCM frames tagged with speaker id.
What else comes with it
Real-time and post-call transcription, webhooks, MIA voice agents, calendar auto-join, and bring-your-own S3 storage.
SDKs cover eight languages. Pair MeetStream with your own model and, if you like, an outside speech engine.
Pricing starts at $0.35/hr and drops to $0.25/hr at volume and for startups. Transcription adds $0.10/hr.
Enterprise-level security, keep your data private.
The certificate, the audit status and the Trust Center are all on the security page.
MeetStream is ISO 27001 certified. SOC 2 Type 2 is under audit rather than complete.
MeetStream is GDPR compliant, and bots run in the United States by default.
Bring-your-own S3 means recordings can live in your storage from the moment they are written.
What the platform carries
Data is encrypted in transit and at rest. The security page has the certificates.
Got a question? We got the answer.
Common questions about capture layers, speech engines and voice-agent frameworks.
What is the difference between a meeting bot API and a speech-to-text API?
A meeting bot API joins the call as a participant and captures the raw audio, video and metadata for you. A speech-to-text API turns audio you already have into text. Most meeting products use both: a bot to capture the media and an engine to transcribe it.
How do I get real-time audio from a Zoom, Google Meet, or Teams meeting?
Use a bot capture API that puts a participant in the call and streams the audio to your server over a WebSocket. With MeetStream, one API call returns real-time audio at roughly 200ms, as binary PCM frames tagged with speaker id, on all three platforms.
Can I build my own AI notetaker on this stack?
Yes. A notetaker is an app built on a capture-and-transcription layer, plus your own summary logic. MeetStream gives you the recording, per-participant audio and speaker-identified transcript from one API. You add the summary and action items on top with your own model.
Is streaming or post-call transcription more accurate?
Post-call transcription is usually more accurate, because the engine has the full audio and can rescore. Streaming trades some accuracy for speed, giving you words within a second or two. Live features need streaming. The record of truth is best taken post-call.
Do I need a voice-agent framework like LiveKit or Pipecat?
You need one when your agent speaks back inside the meeting. Frameworks like LiveKit, Pipecat, and Vapi run the listen-think-speak loop. For an agent that reads and analyses meeting data, a capture API and a model are enough, and MeetStream ships voice agents built in.
How do I get audio separated by speaker?
Use a capture layer that returns a stream per participant. MeetStream gives you per-participant audio and video, plus speaker-identified transcripts, so you can tie every line to the person who spoke it.
Put your agent in the meeting.
One API joins the call, returns speaker-identified media, and streams voice both ways.
Build it on MeetStream
One API to join, record, stream, and transcribe meetings across Zoom, Google Meet, and Microsoft Teams.