Ship AI Agents That Speak In Meetings
MeetMind is the developer SDK for building voice-enabled AI participants that join Zoom and Google Meet. Not to record, but to actually participate
<500ms
Response latency
Relevance engine per turn
<10%
Word error rate
Transcription accuracy
99.9%
Uptime SLA
During business hours
<5min
SDK setup time
Install to first agent
30s
Auto-save interval
Notes synced continuously
How It works
Four Steps.
One Working Agent.
Step 01
Inject context before the call.
Upload documents, agendas, scorecards, and role descriptions. The agent arrives at the meeting fully briefed not cold.
Step 02
Agent joins the call
MeetMind connects to Zoom or Google Meet as an active, named participant, not a bot in the corner. It introduces itself and begins listening immediately.
Step 03
Listens, decides, speaks
The relevance engine scores every speaker turn. When the threshold is met it speaks, asking follow-ups, flagging gaps, holding the conversation naturally.
Step 04
Structured output, immediately
Session ends and a complete scorecard, AI summary, and queryable transcript are available instantly. No manual notes. No memory-dependent write-ups.
Quick Start
Up and running in 60 seconds
Three steps to a live AI agent...
01
Inject Context
Provide agenda, documents, or candidate data before the session starts.
02
Connect to a meeting
Point the adapter at a Zoom or Google Meet link and run.
03
Get structured output
Scorecard, summary, and queryable transcript ready instantly.
Everything to ship a voice agent
The SDK handles audio, context, relevance, and output. You build the product logic on top.
Platform adapters
Plug-and-play connectors for Zoom and Google Meet. No custom connector needed.
Context injection
Feed the agent documents, scorecards, and agendas before the call. It arrives informed.
Live voice participation
The agent joins as a named participant, interacting with audio in real-time.
Relevance engine
Decides when to speak passive, standard, or proactive modes with config thresholds.
Structured note capture
Auto-saves decisions, action items, and open questions every 30 seconds.
Post-session query
Natural language search over the session record. Ask anything, get a direct answer.
Your Agent. Your Rules
Every behaviour is configurable. Set the tone, mode, verbosity, and interruption threshold to match your exact use case.
Every parameter maps directly to the Persona() constructor. Changes to tone, mode, and verbosity affect how the agent speaks — not just when.
Interruption threshold controls how long the agent waits for a natural pause before speaking. Lower values feel conversational. Higher values suit formal sessions.
persona = Persona(name="John",
tone="professional",
mode="standard",
verbosity="medium",
interruption_pause="3s",
auto_record=True,
handle_unknown=False,)Start free. Scale when ready
The core SDK is open source and free forever. Managed infrastructure when you need it.
$0
Open source self-hosted
- Full SDK access
- Zoom & Google Meet adapters
- Unlimited local sessions
- Community Support
$89
Managed API. no infra needed
- Everything in free
- Managed cloud API
- Higher rate limits
- Priority Support
Custom
For large organizations with advanced security and compliance needs.
- Everything in Pro
- Private deployment
- 99.9 uptime SLA
- Dedicated onboarding