Google pushes agent plumbing
Reports and developer posts show Google is moving multi‑agent plumbing into practical surfaces: leaks suggest a Gemini 'Agent' multi‑agent system is being tested and Google published a how‑to for a small multi‑agent blogging system built with its Agent Development Kit. At the same time Google is expanding Gemini Personal Intelligence to more markets like India, showing the company is pairing deeper orchestration with account‑linked personal context. (nokiapoweruser.com) (brianchappell.com) (androidauthority.com) (techcrunch.com)
Google is moving its artificial intelligence agents from developer demos into Gemini itself, while expanding a separate feature that lets Gemini use your own Google data. (developers.googleblog.com) An artificial intelligence agent is software that can plan and take steps toward a goal, not just answer a prompt. Google’s Gemini Agent page says the tool can browse the web, use Google apps, and handle multi-step tasks while asking for confirmation before actions like sending an email or making a purchase. (gemini.google) Google put the developer plumbing in place on April 9, 2025, when it introduced the open-source Agent Development Kit at Google Cloud Next 2025. Google said the kit is “multi-agent by design,” with support for hierarchies of specialized agents, workflow patterns like sequential and parallel steps, and deployment tools for production use. (developers.googleblog.com) The company has kept adding examples. On March 18, 2026, a Google Cloud developer relations post walked through “Dev Signal,” a blogging system built with the Agent Development Kit that uses a root orchestrator, specialized agents, long-term memory, Reddit trend discovery, official documentation lookup, and deployment on Cloud Run. (cloud.google.com) At the same time, Google is shipping more personal context into Gemini. Google India said on April 14, 2026 that Personal Intelligence in the Gemini app is launching in India, with opt-in links to Gmail, Photos, YouTube, and Search so Gemini can answer questions using a person’s own accounts. (blog.google) TechCrunch reported the India rollout starts with Google AI Pro and Google AI Ultra users, with free-user expansion planned “in the coming weeks.” The same report said Google first introduced Personal Intelligence in beta in the United States in January 2026, made it available to all United States users in March, and has also launched it in Canada. (techcrunch.com) Google’s own description shows how the two tracks fit together. Gemini Agent is built to execute tasks across apps and the web, while Personal Intelligence is built to retrieve and reason over a user’s own emails, photos, videos, and searches after the user turns it on. (gemini.google) (blog.google) Google is also drawing a line between automation and control. The Gemini Agent page says the feature is experimental, limited on the web to Google AI Ultra subscribers in the United States using English, and unavailable to Workspace and Student accounts for now. (gemini.google) Google is making a similar argument on privacy for Personal Intelligence. Its India launch post says account linking is off by default, users choose which apps to connect, answers are meant to reference the sources they used, and Gemini has guardrails around sensitive topics such as health. (blog.google) The practical shift is that Google now has all three layers in public view: a framework for building teams of agents, a consumer-facing Gemini Agent that can carry out multi-step work, and a Personal Intelligence layer that can feed Gemini account-linked context. The remaining question is how fast Google broadens access beyond today’s limited markets, subscriptions, and account types. (developers.googleblog.com) (gemini.google) (blog.google)
