Changelog

2026-07-12 — Human calendar connections

Google Calendar and Microsoft Outlook can now contribute human availability to agents. A human authorizes a provider in the Chronary console, selects the calendars that contribute opaque busy time, and grants access to selected agents, all current agents, or current and future agents. Availability queries combine authorized human busy intervals with the agent’s Chronary calendars without exposing OAuth tokens, provider calendar identifiers, calendar names, or private event details to agents.

Humans can optionally publish confirmed Chronary events back to one chosen writable Microsoft Outlook calendar (Google connections are availability-only). Agents may request short-lived setup links through the provider-neutral connection API, while humans can also create and manage connections directly in the console. The bridge is currently available across Free, Pro, and Custom plans.

2026-06-02 — Event reminders

Events can now fire timed reminders. Set a reminders array on an event (minutes before start_time, e.g. [10, 1440] for 10 minutes and 1 day before — max 5 entries, each 1–40320 minutes). Calendars carry a default_reminders field that events inherit when they don’t set their own; resolution falls back to a system default of [10]. A null field inherits the level above, while [] means explicitly no reminders. Each reminder fires a new event.reminder webhook at start_time minus the offset (confirmed events only), and reminders are emitted as VALARM components in the iCal feed so calendar apps display the alarm.

2026-05-11 — Simplified API keys

Removed test mode. Keys no longer distinguish between live and test. Each org has a single API key prefix (chr_sk_*); all requests count toward quotas uniformly. Legacy chr_sk_live_* keys continue to authenticate without change. The mode field has been removed from POST /v1/auth/keys and the is_test field has been removed from all resource schemas.

v1 — Agent-native scheduling

April 16, 2026

v1 extends the Chronary API for true agent-to-agent coordination: agents can now advertise operational state on their calendars, query a single endpoint for temporal context, declare availability rules without placeholder events, and negotiate meeting times through a first-class proposal protocol.

Added

• Temporal context — GET /v1/calendars/:id/context returns the current event, next upcoming event, the last three ended events, the next events starting within 24 hours, and the calendar’s agent_status in a single call
• Agent status — calendars now expose an agent_status field (idle / working / waiting / error) so agents can advertise their operational state to other agents, dashboards, and human subscribers
• Scheduling proposals — full multi-agent negotiation protocol under /v1/scheduling/proposals with create, respond (accept / decline / counter), resolve, and cancel endpoints; resolved proposals emit a confirmed event (spr_ prefix IDs)
• Availability rules — per-calendar PUT / GET / DELETE /v1/calendars/:id/availability-rules for working hours and buffer times, applied automatically by the availability engine
• Temporal holds — tentative event reservations via POST /v1/calendars/:cal_id/events with status=hold, plus PUT /v1/events/:id/confirm and PUT /v1/events/:id/release for promoting or releasing a held slot. Holds carry a TTL (hold_expires_at) and a priority (hold_priority) that pre-empts lower-priority overlapping holds
• Lifecycle webhooks — new event.started, event.ended, event.hold_created, event.hold_expired, event.hold_released, event.hold_confirmed, proposal.created, proposal.responded, proposal.confirmed, proposal.expired, and proposal.cancelled event types for real-time coordination

v1.0 — Public Beta Launch

April 4, 2026

The initial public beta of Chronary, a calendar-as-a-service platform purpose-built for AI agents.

Core API

• Agent CRUD — Create, read, update, delete, and list agents with agt_ prefixed IDs
• Calendar CRUD — Full lifecycle management for calendars (cal_ IDs), linked to agents
• Event CRUD — Create, read, update, delete, and list events (evt_ IDs) with support for single and recurring events
• Availability engine — Query free/busy slots across one or more agents with configurable duration, buffer times, and working hours
• Usage endpoint — GET /v1/usage returns current consumption against plan limits

Integrations

• iCal feeds — Generate read-only iCal feed URLs for any calendar, consumable by Google Calendar, Apple Calendar, Outlook, and other calendar apps
• iCal subscriptions — Subscribe calendars to external iCal URLs; events sync automatically on a recurring schedule
• Webhooks — Subscribe to calendar and event lifecycle events with HMAC-SHA256 signed payloads and automatic retry with exponential backoff
• MCP server — Model Context Protocol server exposing 9 tools, sharing the same Worker and service layer as the REST API

Platform

• API key authentication — Keys use the chr_sk_* prefix, hashed with SHA-256 before storage
• Rate limiting — 10 requests per second per key with Retry-After headers
• Monthly quotas — Plan-based limits on agents, calendars, events, API calls, webhook deliveries, and availability queries
• Console dashboard — Web UI at console.chronary.ai for account management, API key generation, and usage monitoring
• Structured error handling — Consistent error response format with machine-readable type field and request_id on every response for debugging

Infrastructure

• Globally distributed — sub-50ms response times worldwide
• Durable, auto-scaling storage — primary data store with connection pooling
• Observability — structured logging and uptime monitoring
• Test coverage — 327 tests with 90%+ line, branch, function, and statement coverage