Google launches Genkit Agents for full-stack conversational AI
Genkit Agents preview simplifies full-stack conversational AI app development Build chat apps with streaming, tools, persistence, and frontend integration
Google has announced Genkit Agents, a preview feature for building fullstack conversational AI apps with the opensource Genkit framework. The Agents API is available in TypeScript and Go, with support for other Genkit languages as well, and is designed to handle message history, tool use, streaming, persistence, and frontend integration through a single interface.
The company says developers can define agents on the server and run them through the same chat interface whether the agent is local or exposed over HTTP. Genkit also supports session stores for servermanaged conversations, including Firestore for production use, as well as clientmanaged state for apps that handle persistence themselves.
The update adds remoteAgent() for JavaScript frontends, streaming responses, builtin human approval flows for sensitive actions, longrunning detached tasks, and delegation across specialist agents. Google also highlighted a new Agent Runner in the Genkit Developer UI for testing conversations, inspecting snapshots, and debugging tool interactions.