TwinsGoogle Calendar
A Google Calendar twin for testing AI agents.
Test agents that find time, book meetings and move them. All 36 Calendar tools on private calendars, so no real invite goes out.
A stateful copy of Google Calendar, built for testing agents.
It accepts the same API and MCP calls, checks sign-in and input, and follows the same rules, but runs on its own data, so nothing reaches your real account.
- Events, instances and recurrence
- Free/busy and conflicts
- Invites, RSVPs and attendees
- Calendars, sharing and subscriptions
- Time zones and all-day events
Every run starts from a company that already has history. We seed:
- A calendar for every person
- A week of meetings, with conflicts
- Shared team calendars
- Rooms and working hours
Use Google Calendar in staging, without using production.
Validate every action
The twin checks what the agent sends the way Google Calendar would: required fields, permissions, rate limits, malformed input. Mistakes fail here, not in your account.
Start from a realistic state
Describe the company and the situation in plain English. Drizz seeds the twin to match, and resets it to that state before every run.
Replay tasks across tools
The Google Calendar twin shares its people, customers and history with the other twins, so a task that touches three tools runs the way it would at work.
Swap the address. Your agent code stays the same.
Replace the live Google Calendar endpoint with the twin’s. Everything your agent already does keeps working, on data that is yours to break.
{
"mcpServers": {
"calendar": {
"type": "http",
"url": "<twin MCP endpoint>",
"headers": { "Authorization": "Bearer <token>" }
}
}
}REST works on the same data. Both are reset with one call.
Connected to the rest of the company.
Calendar shares its people with Gmail and Slack. An invite the agent sends arrives in the recipient’s mailbox, and the meeting is linked to the deal in Salesforce or HubSpot.
Read the guide: How to test AI agents before production →
17 of 20 runs passed. 3 broke a rule.