ADK

Bring your ADK agents to your users

Give your ADK agents real user-interactivity using CopilotKit and AG-UI. Build rich, interactive, agent-powered applications.

Quickstart
What you can build

Build with ADK

The user-facing primitives every ADK integration ships with — pick the one that fits your product and drop the code in.

Generative UI

Render your agent's state, progress, outputs, and tool calls with custom UI components in real-time. Bridges the gap between AI agents and user interfaces.

Human in the Loop

Empower users to guide agents at key checkpoints. Combine the best of AI and human judgment for more reliable and controllable agent behavior.

Shared State

Keep your agent and your app in sync. Your agent can see everything in your app, and your app can react to your agent in real-time.

Already have Google ADK sessions?

When you add a user-facing app to an existing ADK agent, your users may need to open and continue sessions they started elsewhere. CopilotKit Intelligence can import persisted ADK sessions as Rich Threads, bringing that history into your CopilotKit app.

The importer supports ADK database session stores and Vertex/Agent Engine session history. In-memory sessions cannot be exported; legacy pickle stores need migration first.

Import history once; future CopilotKit-mediated runs synchronize with Intelligence and continue through ADK's native persistence path when your agent uses a durable session service with appropriate retention. This is not a continuous mirror of runs made outside CopilotKit.

Import and synchronize Google ADK sessions for source setup, agent mapping, and thread continuity.

How it fits together

Architecture

The shape of a CopilotKit + ADK application — from your UI down to the agent runtime.

CopilotKit ADK architecture diagram
Live example

Run ADK in your browser

Two patterns we see most often — drive a SaaS workflow, or collaborate on a canvas with your agent.

Render your agent's state, progress, outputs, and tool calls with custom UI components in real-time. Bridges the gap between AI agents and user interfaces.