How Reclaim Evolved Its Calendar Assistant to Be AI-Native
Integrating natural language scheduling while preserving core calendar logic
1Calendar as workflow hub
Reclaim treats calendars as more than time containers - they reflect planning behaviors, coordination needs, and personal priorities. The AI assistant currently handles clear scheduling inputs well but struggles with ambiguous natural language requests. This gap became apparent as LLMs advanced, enabling users to describe scheduling goals conversationally rather than through rigid commands. The challenge was interpreting these requests while maintaining calendar integrity.
2Dual-path scheduling system
Before AI integration, Reclaim operated through two primary mechanisms: direct user edits and automated background scheduling. The automated system considers preferences/availability when allocating time for tasks, focus work, and routines. It dynamically adjusts these blocks as schedules change. This established logic became the foundation for the new AI path, ensuring continuity rather than creating a disjointed experience. All scheduling actors must respect the same core rules to maintain calendar coherence.
3Agent platform architecture
Reclaim's solution was an agent platform that connects LLMs to calendar context without full access. The platform provides models with relevant calendar details and controlled feature access through a tool system. An agent loop manages the interpretation process: the model suggests actions, the agent provides context/tools, and the loop continues until resolution. This architecture prevents LLMs from making arbitrary changes while leveraging their language understanding. Specialized subagents handle complex request components.
4Model Context Protocol
The platform implements Model Context Protocol (MCP) for extended functionality. Internally, MCP lets agents use compatible tools from other services. Externally, it exposes select Reclaim tools to AI clients like Claude and ChatGPT through an MCP server. This allows third-party agents to leverage Reclaim's scheduling capabilities while maintaining security boundaries. The team chose custom implementation over frameworks to better align with their specific scheduling needs and provider APIs.
5Schedule Action standardization
All scheduling operations - whether from users, automation, or AI - convert into standardized Schedule Actions. These internal representations ensure consistent validation and commit processes regardless of origin. This shared path is critical because calendar changes often have cascading effects. Moving one meeting might adjust availability or require rescheduling flexible blocks. Centralizing these operations prevents rule duplication across interfaces as the product evolves.
6Preview Mode safeguards
The team introduced Preview Mode to let users review AI-proposed changes before applying them. This temporary calendar view shows how suggestions would affect existing schedules and shared events. To enable real-time preview adjustments, engineers reworked the scheduler as a pure function that calculates proposals without saving changes. Redis caching ensures fast performance even with complex calendars. Attendee availability updates dynamically within the preview environment.
7Workflow integration principles
The implementation revealed three key requirements for useful AI workflow integration: 1) models need relevant contextual information, 2) actions require controlled execution pathways, and 3) changes need human review before finalization. Reclaim's solution satisfies these by combining the agent platform (context), Schedule Actions (control), and Preview Mode (review). This creates a coherent path from conversational request to implemented calendar change.
8Future-proof architecture
The rebuild positions Reclaim to evolve alongside AI advancements without fragmenting the product experience. New capabilities can plug into the existing scheduling foundation rather than requiring separate implementations. This is particularly valuable as user expectations for natural language interfaces grow. The architecture also accommodates different model providers, ensuring flexibility as the AI landscape changes while maintaining consistent calendar management behaviors.
Original article: Dropbox Tech — How Reclaim Evolved Its Calendar Assistant to Be AI-Native