Open Source · Scheduling Infrastructure · MCP + A2A
Add autonomous scheduling to any AI agent
Open scheduling infrastructure that gives any AI agent autonomous scheduling capabilities — unified availability across all calendars, atomic booking, and zero double-bookings.
For Individuals
Your AI agent manages your calendar
Connect Google, Outlook, and iCloud. Your AI agent — Claude Desktop, Cursor, OpenClaw, Manus, or any MCP-compatible agent — sees your full availability and schedules autonomously.
- ✓ No more double-bookings across calendars
- ✓ Share availability without email ping-pong
- ✓ Works with any MCP-compatible AI agent
For Developers
Add autonomous scheduling to your product
Embed scheduling into your AI product or SaaS — REST API, MCP tools, OAuth for managing user calendars, and framework integrations for LangGraph, CrewAI, and OpenAI Agents.
- ✓ Atomic booking prevents double-bookings
- ✓ 100 bookings/month free, prepaid bundles from $29
- ✓ Budget controls and usage alerts
Why do AI agents keep double-booking your calendar?
Even the latest LLMs — GPT-5, Claude, Gemini — score below 50% on temporal reasoning tasks (OOLONG benchmark). Earlier models scored as low as 29% on scheduling and 13% on duration calculations (Test of Time, ICLR 2025). Most calendar tools for AI agents pass these failures through because they are thin CRUD wrappers — no locking, no conflict detection, and no availability computation. Temporal Cortex solves this with atomic booking via Two-Phase Commit, deterministic RRULE expansion, and cross-provider availability merging.
<50%
LLM temporal reasoning accuracy
Even GPT-5 scores below 50% on temporal reasoning tasks. (OOLONG, Test of Time)
0
MCP servers with locking
Most calendar tools for AI agents are thin CRUD wrappers — no locking, no conflict detection, no availability computation.
3+
Calendar silos per person
Your calendars are fragmented — Google for work, Outlook for enterprise, iCloud for personal. Agents inherit this blindness.
Why build on Temporal Cortex instead of doing it yourself?
Adding scheduling to an AI agent means OAuth flows, provider quirks, timezone edge cases, recurring event math, and race conditions. Temporal Cortex handles all of it — atomic booking with Two-Phase Commit, deterministic RRULE expansion via Truth Engine, and cross-provider availability merging across Google, Outlook, and CalDAV.
Atomic Booking (2PC)
Lock → verify → write → release. If any step fails, everything rolls back. Zero double-bookings, even with concurrent agents. Local mode uses in-process locking. Platform mode uses distributed Redis Redlock for multi-agent coordination.
book_slot({
calendar_id: "primary",
title: "Team Sync",
start: "2026-02-26T14:00",
end: "2026-02-26T15:00"
})
// Lock acquired → conflict check → booked ✓Truth Engine
Deterministic RRULE expansion. Handles DST transitions, BYSETPOS, EXDATE with timezones, leap years. Not API-dependent — computed locally.
expand_rrule({
rrule: "FREQ=WEEKLY;BYDAY=MO,WE,FR",
dtstart: "2026-02-01T09:00",
count: 10
})
// 10 exact occurrences, DST-safe ✓Availability Merging
Returns "these slots are free across all calendars" — not raw events for you to figure out. Google + Outlook + iCloud in one view.
get_availability({
calendars: ["google/primary",
"outlook/work", "caldav/personal"],
date: "2026-02-27"
})
// Merged free slots across all ✓Token-efficient by default — TOON format reduces calendar data by ~40%
All data tools return results in TOON (Token-Optimized Object Notation), a structured format designed for LLM consumption. Fewer tokens per response means faster agent reasoning and lower API costs — without sacrificing readability.
What happens when an LLM expands a recurring event rule?
We ran the same RRULE through frontier LLMs and the Truth Engine. Here's one example — "every other week on Tuesday and Thursday, 8 occurrences":
LLM prediction
FREQ=WEEKLY;INTERVAL=2;BYDAY=TU,TH;COUNT=8
2026-03-03 (Tue)
2026-03-05 (Thu)
2026-03-10 (Tue) ← wrong: should skip week 2
2026-03-12 (Thu) ← wrong: should skip week 2
2026-03-17 (Tue)
2026-03-19 (Thu)
2026-03-24 (Tue) ← wrong: should skip week 4
2026-03-26 (Thu) ← wrong: should skip week 4INTERVAL=2 ignored — generates every week instead of every other week. 4 out of 8 dates wrong.
Truth Engine computation
FREQ=WEEKLY;INTERVAL=2;BYDAY=TU,TH;COUNT=8
2026-03-03 (Tue, week 1)
2026-03-05 (Thu, week 1)
2026-03-17 (Tue, week 3) ← week 2 skipped
2026-03-19 (Thu, week 3)
2026-03-31 (Tue, week 5) ← week 4 skipped
2026-04-02 (Thu, week 5)
2026-04-14 (Tue, week 7) ← week 6 skipped
2026-04-16 (Thu, week 7)INTERVAL=2 applied correctly to the week frequency. Deterministic, RFC 5545 compliant.
How does Temporal Cortex work?
Install with npx @temporal-cortex/cortex-mcp. Authenticate with your calendar provider via OAuth. Your AI agent immediately gets 18 tools across 5 layers (0–4) — discovery, temporal context, calendar operations, availability computation, and booking. The typical workflow follows 4 steps:
Install
npx @temporal-cortex/cortex-mcp
Authenticate
npx @temporal-cortex/cortex-mcp auth google
OAuth opens in your browser. One-time consent, credentials stored locally.
Your agent gets 18 tools across 5 layers
Temporal context, calendar operations, availability computation, safe booking, and open scheduling.
Agent schedules intelligently
get_temporal_context → resolve_datetime → find_free_slots → book_slot
10 seconds to first tool call. Full calendar access within 2 minutes.
Approaches to scheduling for AI agents
When adding scheduling to your AI agent, you have four options — each solves a different problem.
Calendar API wrappers
Nylas, Cronofy
Normalize calendar CRUD across providers. Mature, compliance-ready. But your agent still needs to compute availability, expand recurring events, convert timezones, and prevent double-bookings.
End-user scheduling tools
Calendly, Cal.com
Built for humans sharing booking links. Not developer infrastructure — you cannot embed them into your AI agent's autonomous workflow.
Closed AI scheduling
Single-vendor AI agents
Autonomous agent-driven scheduling tied to a single vendor. Each product works only with its own AI agent — cannot be embedded into your application or combined with other AI agents.
Open scheduling infrastructure
Temporal Cortex
The infrastructure layer between calendar APIs and your AI agent. Handles the hard parts — temporal computation, availability merging, atomic booking — over open protocols (MCP, A2A, REST).
Start free. Scale when you're ready.
Open source with no limits. Free managed platform. Pricing for individuals and developers.
Individual
Pro
$19/mo
- ✓ Unlimited bookings
- ✓ All features included
- ✓ Priority support
Open Source
Free
Forever. No limits.
- ✓ All 18 tools
- ✓ No account required
- ✓ Your machine, your data
npx @temporal-cortex/cortex-mcpThe future is agent-to-agent scheduling
Your agent discovers the best scheduling path automatically. If the other person has an agent, negotiate directly over A2A. If not, compose an email proposal. Temporal Cortex is building the open infrastructure for this transition — from MCP tools to Agent Cards to the agentic web.
Today
The Tool Era
Agents use MCP tools to access calendars. You install a server and your agent gets 18 tools.
2027
The Agent Card Era
Each user gets an A2A-compliant scheduling agent. Agent Cards declare capabilities. Discovery via identity resolution.
2028+
The Agentic Web
Agent-to-agent scheduling in under 2 seconds, zero human intervention. Temporal Cortex becomes DNS for Human Time.
Add scheduling to your AI agent in 10 seconds
Open source. No account required. 18 tools, 5 layers, 4 protocols.
Questions? GitHub Issues are open.