💡 Calendar-Aware Proactive Agent: Meeting-Timed Workflow Suggestions via MCP Calendar Integration

[github-actions[bot]]

Summary

TalkTerm reads the user's calendar via MCP (Google Calendar, Outlook) and proactively suggests contextually relevant workflows timed to their schedule. Before a product review, the avatar offers to prepare a status summary. Before a brainstorming session, it suggests pre-generating idea clusters. This transforms TalkTerm from a tool users must remember to open into a proactive assistant that anticipates their needs.

Market Signal

Microsoft Copilot validates the calendar-aware AI assistant pattern in Teams/Outlook — proactive meeting prep is one of its highest-rated features. Apple's rebuilt Siri AI (WWDC June 2026) demonstrated multi-step task chaining across apps, including calendar-triggered workflows. Enterprise voice AI adoption tripled in 2026. However, no desktop agentic tool combines calendar awareness with local workflow execution for non-technical users. ChatGPT Voice Desktop (July 2026) and Hermes Agent Desktop (June 2026) can both execute agent workflows but neither integrates with calendars.

User Signal

• Existing ideas 💡 Proactive Workspace Monitoring: Avatar-Initiated Project Status Briefings #234 (Proactive Workspace Monitoring) and 💡 Cross-Workspace Morning Synthesis Brief: Avatar-Delivered Daily Digest Across Connected Projects #415 (Cross-Workspace Morning Synthesis Brief) cover project-level proactivity and morning briefings, but neither is calendar-driven
• Calendar-aware suggestions add a TIME-SENSITIVE dimension: the right workflow at the right moment, triggered by the user's actual schedule rather than always-on monitoring
• The distinction matters for non-technical users: "TalkTerm reminded me to prep for my meeting" is a concrete, memorable value moment that drives daily habit formation

Technical Opportunity

• MCP server ecosystem includes Google Calendar and Outlook connectors
• TalkTerm's MCP tool invocation (FR14) and the MCP 2026-07-28 stateless protocol core make connecting to calendar services straightforward — no sticky sessions needed
• Suggestion rules engine can be implemented as a lightweight service in src/main/agent/ that polls calendar events and maps them to workflow templates
• CalendarEvent domain type and SuggestionRule interface in src/shared/types/
• Conversational OAuth for calendar connection (aligns with Discussion 💡 Conversational OAuth: Avatar-Guided MCP Service Connections #293)

Assessment

Dimension	Score	Rationale
Feasibility	med	Requires MCP calendar server integration, event polling, and suggestion mapping; OAuth setup for calendar access adds UX complexity
Impact	med	Differentiates from all desktop agent competitors; drives daily usage habit; but narrower use case than core agent features
Urgency	med	Microsoft Copilot is establishing the pattern; worth tracking but not blocking other work

Adversarial Review

Strongest objection: Privacy concerns with reading calendar data are significant. Users may find proactive suggestions intrusive or annoying, especially if poorly timed.

Rebuttal: Calendar access must be opt-in and clearly explained by the avatar during setup (following the Conversational OAuth pattern from Discussion #293). Suggestion frequency should be conservative: maximum one suggestion per upcoming meeting, never during focus time blocks. Users can mute suggestions or disconnect the calendar at any time via conversation. The "personal assistant" value proposition outweighs the intrusiveness risk when well-calibrated.

Suggested Next Step

Research available Google Calendar and Outlook MCP servers for compatibility with MCP 2026-07-28 spec. Design a CalendarEvent domain type and a SuggestionRule interface in src/shared/types/. Prototype the avatar suggestion flow with mock calendar data.