Skip to content

SYSTEM Cited by 1 source

Reclaim

Reclaim (reclaim.ai) is the Dropbox-owned AI calendar assistant that uses AI to find and adjust time for tasks, habits, and meetings. It works well when inputs and expected outcomes are clear (an automated scheduler places focus work, recurring routines like lunch, and adjusts blocks as schedules change); in 2026 Dropbox extended it with a natural-language interface — describe a scheduling goal in your own words — without rewriting the scheduling core it already had. (Source: sources/2026-09-29-dropbox-evolving-our-calendar-assistant-reclaim-to-be-ai-native-with-f60d451f)

The three actors

Reclaim's central design idea is that three different actors drive the same scheduling system:

  1. The user making explicit changes (move an event, change a setting).
  2. The automated scheduler running in the background — considers each person's preferences and availability to place tasks/focus work/recurring routines and re-adjust them as schedules shift.
  3. The agent (added in the AI-native redesign) — interprets a free-form request, looks at the relevant calendar, and works out how Reclaim can help.

The agent is deliberately a third way to drive the existing system, not a parallel AI-only product. Building a separate AI path "would have duplicated functionality and made Reclaim's behavior harder to keep consistent across interfaces." The hard new property the agent introduces is interpretive variability: an open-ended request like "make time for this tomorrow" can be read as find an open block, create an event, or move something existing — each affecting the calendar (and other people's calendars) differently. That is precisely why all three actors must share the same logic. (concepts/separation-of-concerns) (Source: sources/2026-09-29-dropbox-evolving-our-calendar-assistant-reclaim-to-be-ai-native-with-f60d451f)

The agent platform

Reclaim built its own agent platform rather than adopting a broad agent framework. Division of labor:

  • The model interprets what the user says and figures out intent.
  • The agent manages the larger process: it supplies Reclaim's instructions, the relevant calendar context, and tools (which let it ask Reclaim to look up information or request a specific action). When a tool returns information, the agent feeds it back to the model to inform the next step — the repeated exchange is the agent loop (concepts/agentic-development-loop).

Key platform decisions:

  • Scoped context and tools per request. Reclaim gives the agent only the calendar information and tools relevant to each request, "rather than access to everything at once" — keeping it focused and limited to appropriate actions (a least-privilege + context-budget discipline).
  • Subagents for complex requests. A complex request can have one specific part handed to a specialized subagent; internal checklists and review steps keep the overall request on track (patterns/specialized-agent-decomposition).
  • Build-your-own loop, connect directly to providers. Reclaim built the agent loop, tools, and context system itself and connects directly to model providers. A framework it first explored "lagged behind the provider APIs we wanted to use and included more structure than Reclaim needed." Owning these parts lets them design around scheduling needs, adopt provider capabilities faster, and swap providers without a rebuild — at the cost of more code to maintain (partly offset by using agents to help maintain it).

MCP — used in both directions

Reclaim's tool system supports MCP two ways (patterns/mcp-as-centralized-integration-proxy):

  • MCP client inside the agent loop — lets Reclaim's agent use compatible tools from other services.
  • MCP server — exposes selected Reclaim tools to supported external AI clients (Claude, ChatGPT). Those clients run their own models and agent loops but call the same Reclaim tools used internally.

Schedule Actions — one operation type for three actors

Reclaim represents each scheduling capability — creating or updating an event, changing an RSVP, finding availability — as a Schedule Action, its internal term for a standard operation. Whether the request comes from a user, the automated scheduler, or an agent, the same Schedule Action type is used, "including the same validation and commit process."

This is load-bearing because a calendar change can affect more than one event: moving a meeting can change someone's availability or force Reclaim to adjust flexible events elsewhere. If each actor had its own implementation, engineers would have to reproduce those rules and keep every version aligned as Reclaim evolves. With a shared Schedule Action, "engineers can change how an operation works in one place rather than updating three separate versions," which keeps Reclaim consistent across interfaces. This is a separation-of-concerns move: the operation (and its correctness rules) is one concern, decoupled from who invoked it. (Source: sources/2026-09-29-dropbox-evolving-our-calendar-assistant-reclaim-to-be-ai-native-with-f60d451f)

Preview Mode — review the ripple before it commits

Consistent handling of a change isn't enough; users still need to see its wider effects. Preview Mode gives users a temporary version of their calendar where they review AI suggestions and chat-requested changes before applying them to the real calendar. They can see how an event or settings change affects the rest of their schedule, and confirm effects on shared events before updates are sent to other attendees — a human-in-the-loop review gate on agent output.

To keep Preview Mode fast:

  • The scheduler was reworked into a pure function — it can compute a proposed schedule without saving anything to the real calendar, and Reclaim re-runs that calculation each time the user adjusts the preview.
  • A working copy lives in Redis for fast retrieval, because busy calendars and multi-attendee meetings require a lot of data.
  • Attendee availability is updated within the preview whenever an event moves, so the next calculation reflects the proposed schedule, not the live one.

(Source: sources/2026-09-29-dropbox-evolving-our-calendar-assistant-reclaim-to-be-ai-native-with-f60d451f)

The generalizable lesson

Reclaim's framing of what AI needs to be useful inside an active workflow: surrounding information to understand a request, a controlled way to act on it (shared Schedule Actions), and a review step before the result moves forward (Preview Mode). The long-term payoff is being able to add new AI capabilities on the same scheduling foundation instead of fragmenting the product into a separate experience each time models improve.

Caveats

  • Qualitative post — no latency/throughput/scale/adoption numbers. Treat mechanism claims (pure-function scheduler, per-request scoping, subagent collaboration) as design-described, not benchmarked.
Last updated · 766 distilled / 2,225 read