AI Agents as Workspace Add-ons Across Gmail, Chat, Calendar

Build and deploy AI agents via Google Workspace add-ons that span Gmail, Chat, Calendar, Drive using Cloud Run endpoints calling Vertex AI for contextual trip planning, support, and automations.

Cross-App Agent Deployment Unifies Workflows

Google Workspace add-ons enable single AI agents to operate across Gmail, Calendar, Drive, Chat, and Docs without losing context. Users access the agent via sidebar in Gmail (e.g., for email-triggered trip planning) or dedicated Chat app, with buttons to switch apps seamlessly. The agent uses Gemini's multimodal capabilities to process text, images, and follow-ups—like extracting Paris trip dates from an email, suggesting Air France flights from New York, checking US citizen visa rules (none needed now, ETIAS required late 2026), and grounding responses with Google Search sources. Administrators install domain-wide, boosting team productivity by pulling email context (subject/body) into prompts automatically.

Impact: One deployment handles investigations, customer email aggregation, or support, reducing app-switching and manual data entry.

Internal and External Use Cases Drive Adoption

Two categories emerge: 2P (second-party, internal) for custom automations like incident investigation agents aggregating in-house data or email responders pulling multi-source info; 3P (third-party) for marketplace apps, such as ServiceNow's virtual agent in Chat for issue resolution or Figma's Chat integration with Meet/Docs for notifications, image previews, and diagram comments without leaving conversations.

Impact: Internal builds cut response times (e.g., faster customer emails); external ones scale to other orgs via marketplaces, like Figma keeping users in Chat flow.

Architecture and Code for Production Agents

Core flow: User interaction in Workspace triggers HTTP endpoint on Cloud Run service, which extracts payload (via Chat API for Chat events), user identity, and context (e.g., selected email via extractEmailContents). Augmented prompt feeds Vertex AI Agent Engine (or similar) for responses. Deployment: Enable Google Chat API in Cloud Console, configure app name/avatar/URL/description, interactive features (join spaces), and Cloud Run endpoint. Test locally, then publish internally or to marketplaces with IT approval.

Code entrypoint inspects event.chat for Chat payloads, handles Gmail context injection—basic web app patterns with docs samples. Full travel agent code/tutorial online.

Impact: Cloud Run + Vertex AI delivers reliable, scalable agents; context injection (email body) ensures accurate outputs like flight/visa handling.

Four Takeaways for Builders

  1. Extend Workspace for productivity across Chat/Gmail/Calendar/Drive/Docs. 2. Target 2P (own org) or 3P (others). 3. Single add-on spans apps with shared state. 4. Deploy via Cloud Run to Vertex AI (or alternatives) for agentic logic.
Video description
Pierrick Voulet shows Martin Omander how to build an add on that functions as a persistent AI agent across the entire Google Workspace ecosystem. Instead of just a siloed chatbot, it can follow you from Gmail to Calendar and into Drive, so you aren't constantly context switching or copying data back and forth between tabs. We dig into the architecture of how to bridge these different surfaces using a single codebase and how to use Cloud Run and Google’s Agent Development Kit to make the agent actually useful for automating tasks. *Key Takeaways:* * One codebase, multiple surfaces: How to deploy to all of Workspace. * Agentic workflows: Moving beyond simple "Q&A" to bots that actually interact with your Workspace data. * Contextual awareness: Keeping the AI relevant to the specific document or email you're looking at. Chapters 0:00 - Intro 1:13 - Demo 3:54 - Use cases 5:16 - Architecture 5:42 - Code walkthrough 7:34 - Takeaways Resources: How to get started → https://goo.gle/3PHbhb5 How to write an agent → https://goo.gle/4v1VQKM Host your applications on Cloud Run → https://goo.gle/4bG7DH1 See the Travel Concierge agent code → https://goo.gle/4bWN0oT Speakers: Martin Omander, Pierrick Voulet Products Mentioned: Google Workspace, Google Cloud

Summarized by x-ai/grok-4.1-fast via openrouter

4839 input / 1368 output tokens in 13489ms

© 2026 Edge