Getting started with Google Agent Development Kit (ADK)

Install Python on Windows by downloading Python via the Python manager, install the current latest CPython version, and verify Python and pip versions in the command prompt.

Download PyCharm for Windows, install it, and complete authentication to launch. Import IntelliJ settings or skip, then create a new Python project and run the main file.

Create your first agent with the Google ADK by setting up a PyCharm project, installing requirements, and testing the time agent via the ADK UI using a Google API key.

Run the agent from the CLI with adk run, using the TimeAgent folder, and optionally set a custom port like 9090 to test current time in Dubai (UTC+4).

Explore how adk tools use plain python functions with docstrings to define purpose, arguments, and usage, then connect them to an agent via a tools list.

Explore tool calling by building an order tracking agent with the track order tool, passing an order ID, and retrieving status and the expected date for order updates.

Demonstrate multi-tool calling in the Google Agent Development Kit by adding track order and cancel order functions. Validate order status before cancellation and process refunds.

Build a real-world clinic booking assistant with Google ADK and SQLite, enabling find doctors by specialty, book appointments, and cancel them, all backed by a database.

Introduces the MCP architecture, an open standard that lets ai models securely talk to external tools and services. Explains how the host, MCP client, transport, and servers work together.

This lecture demonstrates connecting an agent to a MySQL MCP server by setting up Ubuntu, creating a demo database, and using ADK web to query employees by department and salary.

Get hands-on with the Google Agent Development Kit (ADK), configuring multiple MCP servers using PostgreSQL and MySQL, connecting via Docker and DataGrip, and querying ecommerce data for project summaries.

Build a custom MCP server with Google ADK and FastMCP to create an inventory system for Ambo KTS, including tools to search products, check stock, and place orders.

Build a university information database using sqlite with an MCP server, create tables, seed data with faker, and implement an interactive chat agent without the adk web UI.

Streamlit provides a shareable web interface for Google ADK agents, wrapping the agent in a Python-based frontend beyond the terminal and ADK web UI.

Discover how Streamlit gives instant UI from Python, built-in charts, reactive execution, session state, and one-click deployment for ADK agents.

Explore practical Streamlit quickstart examples, from a hello world app to user input, simple calculator, data frame filtering, and interactive charts.

Convert the previous MCP server into a Streamlit-based sqlite agent within a Streamlit app; explore page config, markdown, sidebar, prompts, session state, and caching.

Explore rag architecture, including indexing, chunking, embeddings, and vector databases, then use augmented prompts and llm generations to answer from your own data.

Develop a rag application using redis vector store with cosine similarity, upload PDF documents via a stream data application, generate embeddings with a sentence transformer, and perform semantic search.

Develop a rag application using redis vector store, adding search and list document tools, and run with a streamlit UI and ADK agent runner to index and query uploaded PDFs.

Explore the A2A protocol in the Google ADK, enabling networked agent-to-agent communication via http json rpc, and learn when to use or avoid it.

Create a2a server and client, expose a remote greeting agent via the a2a protocol, and test with http requests using uvicorn to access greeting, weather, and time tools.

Explore the core A2A building blocks: agent card, executor, and skills. Understand task lifecycle from submission to completion and how the A2A server and remote agent enable JSON RPC 2.0.

deploy a2a agent on cloud run by building and pushing a docker image, then configure cloud run with app url and api key for library book search and calendar events.

Create and test a cloud run a2a agent with a client using Google ADK, configure calendar and library agents, run ADK web, and troubleshoot prompts and billing session limits.

Explore how the Google Agent Development Kit (ADK) handles memory across short-term, persistent, and long-term memory with Vertex AI Memory, and distinguishes sessions from memory for continuity, personalization, and scalability.

Use the In-Memory Session Service for local development with a simple, zero-configuration setup that stores session data in ram. Avoid persistence; ideal for tests and short-lived stateless tasks.

Explore a short term memory example in Google Agent Development Kit, showing how a memory agent recalls user preferences within a session and differentiates remembered versus lost memory across sessions.

Harness persistent memory with the database session service to store conversations in a relational database via a SQLAlchemy-compatible backend, enabling session retrieval by ID and seamless resume across restarts.

Develop a persistent memory system with Google ADK and SQLite, recalling and saving user preferences by id across restarts via a travel agent app.

Explore Vertex AI Session Service as a cloud managed, memory-integrated solution that scales automatically, preserves cross-session recall, and enables production-grade enterprise security and compliance with minimal code changes.

learn how to deploy a reasoning engine by creating and deploying a travel agent with the Google ADK, configuring Vertex AI session service, project ID, and location.

Update the agent to use Vertex AI session service, remove SQLite code, configure project ID and location from environment, and store memory in Vertex AI session service for long-term memory.

Explore workflow agents in Google's ADK that orchestrate sub-agents with logic for deterministic execution. Discover sequential, loop, and parallel types and how sub-agents can be LLM agents for structured tasks.

Explore the sequential agent pattern in Google ADK, chaining write, review, and refactor in order. See how output keys in state drive the next step via curly brace references.

Build a sequential email processing pipeline with an email classifier, sentiment analyzer, response drafter, and quality checker using Google ADK.

The loop agent enables iterative refinement by running a draft writer and a quality check in sequence, looping up to five times or until the quality threshold is met.

Create a loop agent example project with Google ADK, featuring a product writer agent and QA feedback with QA auditor checks and an exit loop controlling four iterations.

Use parallel agents to run concurrent sub-tasks with fan out, fan in, each on its own session state, returning output keys for downstream synthesis.

Explore a parallel workflow with the Google ADK by running flight, hotel, and attraction researchers in parallel, then synthesize a final trip plan in a sequential pipeline.