Invite any of your agents

Send your agents to meetings

Invite any of your agents. Send your agents to meetings with one API request, and they join Zoom, Google Meet and Microsoft Teams as real participants that wave hello, speak, post in chat, present on camera and call your tools.

Give it any faceBring your own

Invite any of your agents

Claude logoChatGPT logoGemini CLI logoGrok logoHermes logoOpenClaw logoMeta Muse logoPerplexity logoManus logoCursor logoYour own runtime logo

Customers using agents already

Foxsense InnovationsReOrcHHRCGFactory BucketMyDayOneConvogenieADORE HomesZime AINNimblyKaizen SoftworksInterConnectaAI4GroupHedgXTabeerRezoomex

A real participant, built your way

Your agent takes a seat in the call under its own name. It hears each person, answers when it is called on, and gets work done before the meeting ends. You choose the model, the voice and the tools.

An AI agent in a meeting taking audio in and sending voice back
One agent in the call, listening to each voice and answering back.
Hears

Live audio and transcripts, with speaker names.

Speaks

Out loud, in the chat, or silent through tools.

Shows

Images or live video on its own camera.

Acts

Tasks, tickets and CRM updates during the call.

Your stack

Your agent, your stack. No web page to host.

Some platforms ask you to build the agent as a web page and pass its URL. Here you own the whole build: any model, any framework, three ways in.

01

Configure it

Choose a model, a voice and tools in the dashboard or over the API. MeetStream runs the agent in the call.

  • OpenAI, Anthropic, Google or xAI models
  • Wake phrase, barge-in and turn taking
  • MCP tools and custom functions
Voice agents
02

Plug in your framework

Take live audio out and send audio back over WebSockets. Run an OpenAI Realtime session, a LiveKit agent, or any audio in, audio out system.

  • Live audio, mixed or per speaker
  • Speak, chat and interrupt commands
  • Any model you run yourself
Bridge guide
03

Let an agent drive

Connect the hosted MCP server, and the assistant you already use sends the bot, reads the call and posts in the chat.

  • 19 tools for bots and transcripts
  • Sign in with OAuth or an API key
  • Works in any MCP client
Developer tools
In the call

Everything a participant can do, your agent can do

Pick a skill to see it in a call. Each one is a setting on the agent.

Live captions in a Google Meet call where Priya asks the agent Atlas about the launch date and Atlas answers out loudThe Google Meet chat panel where Daniel asks the agent for the action items and Atlas replies with a numbered listA Google Meet call where the agent tile shows a live summary of decisions and action items with ownersThe Google Meet chat panel where the agent confirms it created four tasks and moved a ticket during the callA Google Meet call next to a panel of five separate live audio tracks, one per participant, with a speaker-tagged transcriptThe Google Meet chat panel where the agent lists the account notes, tickets and agenda it loaded for this call, then answers a question from them
Reply modeVoice
Wake phraseIts name

Priya says the agent's name, and the agent answers out loud. Illustration of a MeetStream agent in a Google Meet call.

On camera

Show the room what it sees

The agent owns its camera tile. Send an image, or stream live video into it: a running summary, your slides or a product demo. On Zoom, live video uses your own Zoom app.

  • An image as its video frame
  • Live video into its camera
  • Slides or a product demo
  • An optional avatar
A Google Meet call where the agent tile shows a live summary of decisions and action items with owners
The agent shows a live summary as its video frame. Illustration of a MeetStream agent in a Google Meet call.
Voice control

The controls that make an agent good company

These decide whether an agent helps a meeting or talks over it. Each one is a setting on voice agents.

  1. Wake phrase
    In pipeline mode the agent waits for a phrase you set, such as its name, then keeps listening for 30 seconds.
  2. Barge-in
    In pipeline mode, talking over the agent stops it, while a short "mm-hm" lets it finish the point.
  3. Turn taking
    A turn detection model, or plain silence, decides when a person has finished speaking.
  4. Reply mode
    Voice, meeting chat, or silent action through tools. Set per agent, and different per meeting if you like.
One request

In the meeting with one request

Send the meeting link and the agent to use. The second object fills that meeting's context into the agent's prompt.

Send an agent to a meetingbash
curl -X POST "https://api.meetstream.ai/api/v1/bots/create_bot" \
  -H "Authorization: Token <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "meeting_link": "<ZOOM_MEET_OR_TEAMS_LINK>",
    "bot_name": "Atlas",
    "agent_config_id": "<YOUR_AGENT_ID>",
    "agent_config_params": { "account": "Northwind", "agenda": "Renewal" }
  }'
Use cases

What teams send into meetings

The same agent runtime powers very different products. See ten use cases in detail.

  1. Live sales coach
    Hears the customer, checks the CRM, and prompts the rep in the chat while the deal is live.
  2. Interview assistant
    Follows a question plan, keeps each voice apart, and scores answers as the interview runs.
  3. Support agent
    Joins an escalated call with the ticket history loaded, and files the follow-ups before it ends.
  4. Meeting teammate
    Answers questions from your docs, tracks decisions, and turns action items into tasks.
Platforms

One request, every platform

The same request joins Zoom, Google Meet and Microsoft Teams. Zoom needs no setup: the meeting link and your API key are enough.

  • Zoom with no setup
  • Google Meet
  • Microsoft Teams
  • WhatsApp calls in Beta
Security

Enterprise-level security, keep your data private.

The certificate, the audit status and the Trust Center are all on the security page.

Your data, your retention

Keep recordings with MeetStream, or write them straight to your own storage bucket.

Engineers on the other end

Support on Slack, email and chat, with a response inside one working day.

One bot or thousands

Concurrent bots scale on their own. There is no cluster for you to run.

SOC 2Type 2 (Under Audit)
ISO 27001Certified
GDPRCompliant
FAQ

Got a question? We got the answer.

Common questions about building agents, models, voice and platforms.

How do I send my agents to meetings?

Create an agent, then send one API request with the meeting link and the agent to use. The bot joins Zoom, Google Meet or Microsoft Teams, and the agent hears the room and replies.

Do I have to build my agent as a web page?

No. There is no page to host and no URL to pass. Configure the agent in MeetStream, connect your own framework over WebSockets, or let an assistant drive it through the MCP server.

Which models and frameworks can I use?

The built-in runtime runs OpenAI, Anthropic, Google and xAI models. With your own pipeline you can run any model, for example an OpenAI Realtime session or a LiveKit agent.

Can the agent speak, chat and show video?

Yes. It can answer out loud, post in the meeting chat, and show an image or live video on its camera. On Zoom, live camera video uses your own Zoom app.

Does Zoom need any setup?

No. Zoom needs only the meeting link and your MeetStream API key. Your own Zoom app is optional and adds signed-in joins and live camera video.

Can I control what the agent knows in each meeting?

Yes. You fill the prompt with that meeting's context when you send the bot, limit its tools to an allow-list, and can change the prompt during the call.

Send your first agent

One request puts it in the call. The stack behind it is yours.

Any model
Any framework
Voice, chat, tools
Zoom, Meet, Teams