Evolving our calendar assistant Reclaim to be AI-native without starting over
Source
Greg Unrein,Josh Jensen,Christopher Wildman
Author
Greg Unrein,Josh Jensen,Christopher Wildman
Date
Key takeaways · AI-distilled
Reclaim sends every calendar change, whether it comes from a user, the background scheduler, or an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →, through one internal operation type called a Schedule Action with shared validation and commit, so scheduling rules change in one place instead of three.
Dropbox built its own agentic loopThe cycle an agent runs in: decide, call a tool, read the result, decide again — repeating until the goal is met or a stop condition fires.Full definition →, tools and context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → gathering and calls model providers directly, after a framework it tried lagged behind provider APIs and imposed more structure than Reclaim needed. The tradeoff it names is more code to maintain.
Each request exposes only the calendar data and tools relevant to it rather than everything at once, and a complex request can hand one part to a specialized multi-agentUsing several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.Full definition →, with internal checklists and review steps keeping the whole on track.
Preview Mode shows agent-proposed changes on a temporary copy of the calendar before attendees see anything. To keep it fast, Dropbox reworked the scheduler as a pure function that computes a schedule without saving, and caches busy calendar data in Redis.
The same tool system supports MCP in both directions: an MCP client lets Reclaim's agent use compatible outside tools, and an MCP server exposes selected Reclaim tools to clients such as Claude and ChatGPT.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
agentic loop — The cycle an agent runs in: decide, call a tool, read the result, decide again — repeating until the goal is met or a stop condition fires.
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
Why it matters
Shows one path for adding conversational agents to a mature product: ground requests in existing calendar context and keep the current scheduler, rather than rebuilding from scratch.
