An Interaction Is All You Need — Ivan Leo, Google DeepMind

Models became agents, and the APIs didn't keep up. Google DeepMind rebuilt them. Ivan Leo, developer experience engineer at Google DeepMind, introduces the Gemini Interactions API and Managed Agents. He traces how we went from single completions to function calling to agents, then shows server-side state with interaction IDs (no more hand-managing thought signatures), strongly typed multimodal outputs, mixing built-in and custom tools, and the new steps data model. He then demos Managed Agents, which run the Antigravity harness in a persistent remote sandbox with environment IDs, loadable sources, a credential-injecting proxy and named agents, plus the open-source Gemini API CLI. In this talk: • Why agent workloads need a different API shape • Server-side state: interaction IDs and thought signatures • Chaining image, video and audio generation with the same context • Managed Agents: persistent sandboxes, skills, secure credentials and named agents SPEAKER Ivan Leo, Developer Experience Engineer, Google DeepMind LinkedIn: https://www.linkedin.com/in/ivanleo X: https://x.com/ivanleomk GitHub: https://github.com/ivanleomk Website: https://ivanleo.com/ LINKS Interactions API overview: https://ai.google.dev/gemini-api/docs/interactions-overview Managed Agents quickstart: https://ai.google.dev/gemini-api/docs/managed-agents-quickstart Antigravity agent: https://ai.google.dev/gemini-api/docs/antigravity-agent Gemini API CLI (GitHub): https://github.com/google-gemini/gemini-api-cli Google AI Studio: https://ai.google.dev/aistudio CHAPTERS 0:00 Intro 0:12 Interactions API and Managed Agents 0:27 From single completions to agents 1:02 Function calling 1:32 Models that reason and act 2:02 What is an agent? 3:02 Try the models in AI Studio 4:07 Why a new API 5:31 Server-side state and thought signatures 6:21 Demo: chaining image and video generation 7:01 How the code works 7:36 Strongly typed outputs 8:21 Mixing built-in and custom tools 9:01 The steps data model 9:46 Managed Agents: Antigravity in the cloud 10:36 Demo: analyzing a GitHub repo 11:06 Persistent sandboxes with environment IDs 11:41 Loading sources: GCS, GitHub, inline files 12:16 A 2M-token repo analysis 12:56 The same agent locally and in the cloud 13:36 A credential-injecting proxy 14:26 Named agents 15:26 The Gemini API CLI 15:56 Migrating with the Interactions API skill 16:36 Wrap-up Recorded at the AI Engineer World's Fair 2026 in San Francisco. Subscribe for more talks from the engineers building with AI. AI Engineer: https://ai.engineer YouTube: https://www.youtube.com/@aiDotEngineer X: https://x.com/aiDotEngineer LinkedIn: https://www.linkedin.com/company/aidotengineer/ #Gemini #AIAgents #AIEngineer