Agentic AI with Google ADK & MCP Servers

Explore agentic AI with the Google SDK, showing how external data sources, orchestration, and model-agnostic, open-source tools enable end-to-end tasks like real-time restaurant reservations.

Explore Google ADK overview, learn to build your first ai agent with a flexible, model-agnostic framework, integrating tools, orchestration, memory, and hands-on setup.

Create your first ADK agent by installing Google SDK, building a Python package with agent.py and init.py, selecting a Gemini model, and configuring prompts and authentication for the web UI.

Learn to extend an agent with a Google search tool by importing the Google SDK, adding the tool, and routing queries to real-time search results via tool calls and traces.

Demonstrates an ADK agent using custom function tools morning grid and evening grid, selecting the appropriate tool based on user greetings to generate mood-based responses.

Explore terminal and API-based agent execution with authenticating options using Vertex AI, API keys, and application default credentials, including dot env setup and local versus production deployments.

Explore how tools in the Google agent SDK empower AI to act beyond text with built-in, function, and third-party tools, plus hands-on demos of GitHub actions and BigQuery queries.

Learn how the root agent uses its reasoning capability to pick and invoke the right tool, like BigQuery data queries or GitHub repo creation.

Explore the types of tools in agentic AI, including function tools, agent tools, long-running tools, built-in tools, and third-party tools, with examples like Google search and BigQuery.

Build a CDK agent with built-in tools, including Google search and BigQuery, to list datasets and tables, describe schemas, run SQL, and read data in read-only mode.

Extend your sdk agent with third-party tools like LangChain Wikipedia tool and Q AI scrapers to research on Wikipedia and scrape Google Cloud release notes for Gemini updates.

Explore function tools in the SDK agent to create or delete GitHub repositories by collecting name, description, and privacy, then invoking GitHub APIs.

Discover the google agent development kit's three agent types—lm agent, workflow agent, and custom agent—and build multi-agent systems with sequential, parallel, and loop workflows, including BigQuery integration.

demonstrates building a sequential workflow using a sequential agent to orchestrate a multi-agent job search: fetch open positions, generate interview questions and answers, then produce interview tips.

Explore a parallel multi-agent system design for content creation, using three sub agents for YouTube, blog, and Instagram to generate parallel outputs via the Google SDK.

Design a loop agent that uses an initial writer, critic, and refiner to iteratively craft a Google Cloud services architecture for a web app, using state and max iterations.

Deploy a multi-agent system to Google Cloud Run with the Google Agent Development Kit. Run a content creation workflow in parallel using blog, YouTube, and Instagram agents.

Deploy the ADK agent to Google's agent engine (now agent runtime) using Vertex AI credentials, test in the playground, and explore traces, logs, and open telemetry for troubleshooting.

Learn how callbacks empower agent ai workflows by inserting before and after checks for agent, model, and tool stages, with demos on prompts, access, and logging.

Validate before agent callback and after agent callback access to guide the agent execution flow. Define check access against a small allowed list and log completion after execution.

Explore model callbacks in the SDK, including before model, after model, and agent callbacks, showing how to append prompts, control outputs, and enforce access rules.

Explore tool callback hands-on, showing before and after tool callbacks to filter HR data by user access and country via a BigQuery MCP integration.

Explore the model context protocol (MCP) to connect LLMs with external systems like BigQuery and Google Maps through MCP servers and clients, including local and remote deployments, tools, and architecture.

learn to create your first MCP server with VS Code using a Google Maps MCP server, configure an API key, enable Copilot agent mode, and test map tools.

Learn to configure MCP servers with Google Maps on a cloud desktop, integrating the Google Maps API via a config.json, and run map search and directions using the MCP servers.

Integrate google maps mcp servers with the sdk agent using the mcp tool set class, enabling real-time tool access like geocoding, places, and directions via local mcp services.

Set up an MCP toolbox for database server and connect it to your SDK agent using BigQuery and connectors. Explore secure tools for sql and nosql databases from the IDE.

Set up the MCP toolbox for the database in VS Code, connect to BigQuery, manage datasets and tables, and run Copilot-assisted SQL using Google credentials.

Install and configure MCP toolbox for database to create MCP servers for BigQuery. Then build an SDK agent and an HR analyst agent to answer natural-language questions from BigQuery data.

Explore context management in ADK, detailing short-term and long-term memory, artifacts, and sessions, and learn how memory stores and retrieves information across agent conversations, with hands-on demos.

Understand how session and state function as short-term memory in agentic systems, storing conversation history and user data across a session, with in-memory, Vertex AI, and Cloud SQL options.

Explore hands-on with ADK to create an HR agent and configure session services from in-memory to SQLite and cloud SQL, tracking session IDs, events, and persistent state in vertex AI.

Explore hands-on state management in agentic workflows, detailing session, user, app, and temporary states, their scopes, persistence, and practical demos like BigQuery output and role-based access.

Explore long-term memory in Google ADK, compare it to short-term memory, and learn when to use in-memory or Vertex AI memory bank for persistent, semantic memory across sessions.

Build an agent with memory using the Google ADK, exploring short-term and long-term memory and persisting it with Vertex AI memory bank.

Explore Vertex AI memory bank, a cloud native persistent memory service that stores selective user facts across sessions to create stateful agents with personalized responses.