google-adk-python

Build multi-agent workflows in Google ADK Python by using SequentialAgent, ParallelAgent, and LoopAgent patterns to orchestrate structured pipelines. This toolkit enables hierarchical coordination with sub-agents and coordinator patterns for predictable AI agent execution.

Deploy AI agents to Vertex AI or Cloud Run using this toolkit's code-first deployment capabilities. It supports scalable deployment across Google Cloud infrastructure, allowing you to move from local multi-agent orchestration directly to production environments.

Yes, you can use custom Python functions as tools for your AI agents. The toolkit supports a tools ecosystem that includes pre-built tools, OpenAPI integration, and custom Python functions to extend agent capabilities within your workflows.

Yes, Google ADK Python supports human-in-the-loop approval processes. It provides a code-first workflow that integrates tooling and safety considerations, allowing teams to model automation pipelines with manual approval steps before execution.

The best way to coordinate multiple sub-agents in an AI pipeline is using hierarchical coordination patterns. Google ADK Python provides coordinator patterns and structured workflow agents to manage complex multi-agent systems with predictable execution.

Use LoopAgent for workflow orchestration when your process requires iterative execution, and ParallelAgent for concurrent task execution. Google ADK Python provides these distinct workflow patterns to structure AI agent pipelines based on your specific execution requirements.