Google ADK Multi-Agent

v1.0.0Coding Agentsstable

This project demonstrates a multi-agent system using Google's Agent Development Kit (ADK), Agent2Agent (A2A) and Model Context Protocol (MCP). that integrates Notion for information retrieval and ElevenLabs for text-to-speech conversion.

adk-mcp-a2a-crash-coursemcpai-integration
Share:
53
Stars
0
Downloads
0
Weekly
0/5

What is Google ADK Multi-Agent?

Google ADK Multi-Agent is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to this project demonstrates a multi-agent system using google's agent development kit (adk), agent2agent (a2a) and model context protocol (mcp). that integrates notion for information retrieval and elev...

This project demonstrates a multi-agent system using Google's Agent Development Kit (ADK), Agent2Agent (A2A) and Model Context Protocol (MCP). that integrates Notion for information retrieval and ElevenLabs for text-to-speech conversion.

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • This project demonstrates a multi-agent system using Google'

Use Cases

Build multi-agent systems with Google's Agent Development Kit. Integrate Notion for retrieval and ElevenLabs for speech.
chongdashu

Maintainer

LicenseMIT
Languagepython
Versionv1.0.0
UpdatedMay 18, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx adk-mcp-a2a-crash-course

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Google ADK Multi-Agent

The ADK MCP A2A Crash Course is a reference multi-agent system demonstrating how to combine Google's Agent Development Kit (ADK), the Agent-to-Agent (A2A) communication protocol, and MCP tools in a single application. It runs three coordinated agents: a HostAgent that orchestrates, a NotionAgent that retrieves information from Notion databases, and an ElevenLabsAgent that converts text to speech — all wired together via A2A over HTTP. Developers use it as a hands-on learning project and architectural template for building their own production multi-agent pipelines with ADK.

Prerequisites

  • Python 3.11+ and the uv package manager installed
  • A Notion account with an integration API key (NOTION_API_KEY) and at least one shared database
  • An ElevenLabs account with an API key (ELEVENLABS_API_KEY) for text-to-speech conversion
  • An Anthropic API key (ANTHROPIC_API_KEY) for the Claude model powering the agents
  • Git to clone the repository
1

Clone the repository

Clone the crash course repository from GitHub.

git clone https://github.com/chongdashu/adk-mcp-a2a-crash-course.git
cd adk-mcp-a2a-crash-course
2

Create and activate a virtual environment

Set up a Python virtual environment using venv or uv.

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
3

Install dependencies

Install the project package with all development dependencies using uv.

uv pip install -e ".[dev]"
4

Configure environment variables

Copy the example environment file and fill in your API keys and agent URLs.

cp .env.example .env
# Edit .env and set:
# NOTION_API_KEY=your_notion_integration_key
# ELEVENLABS_API_KEY=your_elevenlabs_api_key
# ANTHROPIC_API_KEY=your_anthropic_api_key
# HOST_AGENT_A2A_URL=http://localhost:8001
# NOTION_AGENT_A2A_URL=http://localhost:8002
# ELEVENLABS_AGENT_A2A_URL=http://localhost:8003
5

Start all agents and the UI

Launch all three agents using the startup script, then start the Streamlit UI to interact with the multi-agent system.

# Start all agents
python scripts/start_agents.py

# In a separate terminal, launch the UI
streamlit run ui/a2a_app.py --server.port 8080
6

Test individual agents

You can also test each agent independently before running the full orchestrated system.

# Test just the Notion agent
python -m notion_agent --port 8002
python notion_agent/test_client.py

Google ADK Multi-Agent Examples

Client configuration

Environment file configuration for the ADK multi-agent system with all required API keys and agent URLs.

{
  "env": {
    "NOTION_API_KEY": "your_notion_integration_api_key",
    "ELEVENLABS_API_KEY": "your_elevenlabs_api_key",
    "ANTHROPIC_API_KEY": "your_anthropic_api_key",
    "HOST_AGENT_A2A_URL": "http://localhost:8001",
    "NOTION_AGENT_A2A_URL": "http://localhost:8002",
    "ELEVENLABS_AGENT_A2A_URL": "http://localhost:8003"
  }
}

Prompts to try

Example prompts to use in the Streamlit UI to interact with the multi-agent system.

- "Find the project notes in Notion and read them to me using text-to-speech"
- "Look up the meeting notes from last week in Notion and summarize them"
- "Get the content from my 'Product Roadmap' Notion database and convert it to audio"
- "Search Notion for information about our API design guidelines"
- "Read aloud the latest entries from my Notion journal database"
- "Find and summarize all action items in my Notion workspace"

Troubleshooting Google ADK Multi-Agent

NotionAgent cannot find or access databases

Notion integrations must be explicitly shared with each database you want to access. In Notion, open the database, click '...' > 'Add connections', and select your integration. The NOTION_API_KEY must belong to an integration that has been granted access to the specific pages or databases you want to query.

Agents fail to start with 'port already in use' error

Another process is using ports 8001, 8002, or 8003. Run 'lsof -i :8001' (macOS/Linux) to identify the conflicting process and kill it, or update the A2A URL environment variables in .env to use different ports and restart the agents.

ElevenLabs audio generation fails or returns empty output

Verify your ELEVENLABS_API_KEY is valid and your account has available character credits. Free tier accounts have monthly limits. Check the ElevenLabs dashboard for quota usage. Also ensure the text being sent for conversion is within ElevenLabs' character limits per request.

Frequently Asked Questions about Google ADK Multi-Agent

What is Google ADK Multi-Agent?

Google ADK Multi-Agent is a Model Context Protocol (MCP) server that this project demonstrates a multi-agent system using google's agent development kit (adk), agent2agent (a2a) and model context protocol (mcp). that integrates notion for information retrieval and elevenlabs for text-to-speech conversion. It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Google ADK Multi-Agent?

Follow the installation instructions on the Google ADK Multi-Agent GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Google ADK Multi-Agent?

Google ADK Multi-Agent works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Google ADK Multi-Agent free to use?

Yes, Google ADK Multi-Agent is open source and available under the MIT license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "adk-mcp-a2a-crash-course": { "command": "npx", "args": ["-y", "adk-mcp-a2a-crash-course"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Google ADK Multi-Agent?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides