Use Case

Eight tools make your AI agent a scheduling assistant

Connect any agent to real calendars over MCP. It checks availability, books, reschedules, and cancels, with no human in the loop. For real.

Get started free MCP docs

Sound familiar?

to schedule a single meeting with a prospect

Most AI agents can draft emails, summarize documents, and write code. But when it comes to actually booking a meeting, they punt the scheduling link to you. You're still the one opening Calendly.

for AI agents in most scheduling platforms

Acuity, SavvyCal, and most booking tools offer no MCP support at all. Calendly shipped a hosted MCP server in March 2026 with a narrower toolset. Without MCP, your agent can't book anything unless you build a custom integration from scratch.

available in Astrocal's MCP server

Check availability, list event types, create bookings, reschedule, cancel, join waitlists, and more. Your AI agent gets the same scheduling capabilities as a human assistant.

How AI agent scheduling works

The booking flow is straightforward. Your AI agent gets a scheduling request like "book a 30-minute demo with the lead" and handles it end to end.

First, the agent calls list_event_types to see what meeting types are available. Then it calls check_availability with a date range to find open slots, which Astrocal calculates against your connected calendars, buffer times, and working hours. The agent picks a slot, calls create_booking with the invitee's name and email, and Astrocal sends confirmation emails and calendar invites automatically.

The entire exchange takes a few seconds. You get a notification via webhook. No link was shared, no back-and-forth happened, and you didn't touch the keyboard.

Rescheduling works the same way. The agent calls check_availability again to find a new time, then reschedule_booking with the booking ID and new start time. The invitee gets notified. The calendar event updates.

npm install @astrocal/mcp-server

# Or run directly with npx (no install required)
npx @astrocal/mcp-server

🔍

Check availability

The agent queries open slots for any event type across a date range, respecting calendar conflicts, buffer times, and your working hours. It only sees what's genuinely free.

📅

Create bookings

Book a meeting directly from the conversation. No link sharing, no waiting for the invitee to find a time. The agent handles the API call and Astrocal sends the calendar invite.

🔄

Reschedule and cancel

Change or cancel bookings via natural language. The agent calls reschedule_booking or cancel_booking, the invitee gets notified automatically, and the calendar event updates.

🌍

Multi-timezone

Pass an IANA timezone when checking availability or creating a booking and Astrocal handles the conversion. The agent doesn't need to calculate offsets manually.

🛡️

Respect your rules

Buffer times, working hours, and max bookings per day are all enforced server-side by the API. The agent can only book into slots that actually exist. No edge cases to handle in the prompt.

🔔

Webhook notifications

Every AI-booked meeting fires a webhook. Push notifications to Slack, update your CRM, or trigger any downstream workflow when an agent creates, reschedules, or cancels a booking.

Which AI agents work with Astrocal?

Any agent that supports MCP works natively. Claude, via Claude Desktop or Claude Code, connects directly using the JSON config above. Restart the app and the tools appear automatically, no code required.

For GPT-based agents and custom agents using the OpenAI function-calling interface, MCP bridges like MCP-to-OpenAPI can expose the same tools. Alternatively, any agent can use the REST API directly. It's the same endpoints the MCP server calls under the hood.

Custom agents built with the MCP SDK can connect to Astrocal the same way they connect to any other tool server. The server runs as a stdio process, so it works with any MCP client implementation regardless of the underlying model.

If your agent doesn't support MCP and you don't want to add a bridge, the REST API and SDKs are always available. The MCP server is a convenience layer, not a requirement.

Frequently asked questions

The Model Context Protocol is an open standard for connecting AI agents to external tools and data sources. It gives language models a structured way to call functions. Astrocal publishes an MCP server that exposes scheduling operations as callable tools: check availability, create bookings, reschedule, and more.

No. The MCP server only exposes availability slots and booking operations. It cannot read calendar event details, attendee lists, meeting subjects, or any other calendar data. The agent sees open time slots, not your calendar contents.

API keys currently grant full API access; scoped read-only keys are not available today. What you can do now: develop against a sandbox test key (ac_test_) so the agent can't touch production data, revoke any key instantly from the dashboard, audit every agent action in the activity log, and watch bookings in real time via webhooks.

The MCP server works natively with Claude (Claude Desktop, Claude Code, and any Claude API client that supports MCP). For GPT-based agents, you can use the Astrocal REST API directly or connect via an MCP bridge. The REST API supports the same operations as the MCP server.

No. The MCP server is included in every Astrocal plan, including the free plan. There's no separate MCP tier or add-on. Install the package, add your API key, and it works.

Give your AI agent a calendar

Install the MCP server and let AI agents book meetings in minutes.

Get started free