AI GitHub
Google ADK for Python agent applications
Create agents and workflows with tools, session state and development utilities.
Key Takeaways
- The Agent Development Kit separates model-driven agents from workflow agents that coordinate steps.
- Choose a model backend and decide how sessions will be created for requests.
- Run a single tool-using agent in the documented development environment, then add a sequential workflow step.
- Providers and deployment targets bring their own configuration.
How it works
The Agent Development Kit separates model-driven agents from workflow agents that coordinate steps. Session and runner abstractions connect the agent definition to application execution.
What you need
Choose a model backend and decide how sessions will be created for requests. Keep persistent state separate from any one developer's local testing environment.
What to try first
Run a single tool-using agent in the documented development environment, then add a sequential workflow step. Compare the visible session state across two independent requests.
Reset a test session and run the same request again. If the results differ, inspect state and tool input before changing the prompt or the model.
Where to look for failures
Providers and deployment targets bring their own configuration. Pin the example's model and isolate sessions when testing data access for several users.
Related Franklin pages
Build the same two-step workflow with one tool and one failure case. Inspect how state, errors and tool permissions cross the boundary between the steps.
- Represent an agent workflow as nodes, transitions and explicit state. LangGraph for stateful agent workflows
- Connect model-backed agents, tools and explicit workflows in Python or .NET. Microsoft Agent Framework for orchestrated agents
Browse Franklin for concrete examples of tools and services exposed through MCP. Franklin’s MCP directory
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗