Details
- Google for Developers announces the Genkit Agents API, a new server-side framework for building agentic, full-stack AI applications.
- The Agents API is released in preview for TypeScript and Go, with support for defining agents on the server and driving them via a unified chat interface.
- Genkit agents package core conversational plumbing — model loop, message history, streaming, tool calling, and persistence — behind a single API, reducing repetitive infrastructure work.
- Developers can add a session store so the server manages conversation history, custom state, and artifacts, enabling clients to resume sessions using a session ID.
- The same chat() interface can be used whether the agent runs locally or behind an HTTP endpoint, simplifying integration from browser or mobile frontends through remoteAgent.
- The Agents API supports HTTP routes, interrupts, background work, and JavaScript clients calling agent endpoints, fitting naturally with multi-step, conversational features.
- Genkit, Google’s open-source framework, already offers SDKs for JavaScript/TypeScript, Go, Python and Dart, and the Agents API extends it with higher-level agentic workflows.
- Google positions the Agents API as a way to configure, rather than rebuild, the plumbing for tool loops, streaming updates, and multi-turn conversations in production AI apps.
Impact
By introducing a higher-level Agents API on top of Genkit, Google is lowering the barrier to building production-grade AI agents that span server and client. This move aligns Genkit more closely with emerging agent frameworks in the AI ecosystem and strengthens Firebase and Cloud Run as default backends for conversational and tool-using AI workloads, pressuring rivals to offer similarly integrated agent tooling.
