androidengineers.Book a session

What an in-app agent does

What an in-app agent does

articleSelf-paced

Understand the boundaries between agent reasoning and Android capabilities.

1. Runtime placement

An Android agent executes orchestration inside the app. Its model may run locally or remotely. Local orchestration alone does not mean the user data stays on-device.

2. Android-specific setup

Use the documented Android ADK artifact and compatible processor configuration. Android and JVM artifacts are alternatives, not dependencies to include together.

3. Capability boundaries

The agent can only use explicitly exposed tools. It does not automatically gain access to contacts, files or other apps. Android permissions and domain checks remain authoritative.

Worked scenario

A task assistant can search app-owned notes through one tool while having no permission to inspect arbitrary device files.

Apply it

Draw the runner, model adapter, tool service and Android permission boundaries.

Check your understanding

You can distinguish local orchestration from local inference. Explain the decision and show evidence from your implementation or design. If you cannot demonstrate it yet, revisit the relevant section before continuing.

YOUR LEARNING JOURNEY

0 of 13 available lessons completed

Progress saved in this browser. No account needed.
What an in-app agent does | ADK for Android | Android Engineers