This directory contains practical examples demonstrating how to use the Python A2A library. Each example is designed to showcase different aspects of the A2A protocol and help you quickly implement agent-to-agent communication in your projects.

If you're new to Python A2A, start with these examples:

1. Hello A2A - Create your first A2A messages in just a few lines of code
2. Simple Client - Connect to any A2A-compatible agent
3. Simple Server - Create your own basic A2A server
4. Basic Streaming - Implement real-time streaming responses

These examples provide a gentle introduction to A2A concepts:

These examples demonstrate how to implement streaming responses:

These examples demonstrate fundamental A2A components:

Connect A2A to various AI services:

Build and manage networks of cooperating agents:

Add external tool capabilities to your agents with the new provider architecture:

End-to-end examples for real-world use cases:

Tools to enhance your development workflow:

• AI/ML Engineers: Start with the AI-powered agent examples to see how to integrate with LLMs
• Backend Developers: Begin with the building blocks examples to understand the core protocol
• Frontend Developers: The simple client example shows how to connect to A2A agents
• DevOps Engineers: Check the CLI tools example for automation capabilities

• I want to call an existing AI agent: Simple Client
• I need to build my own agent: Simple Server or Agent Skills
• I want to use OpenAI with A2A: OpenAI Agent
• I need to implement streaming responses: Basic Streaming
• I want to integrate streaming with LLMs: LLM Integration
• I need to build a streaming UI: UI Integration
• I need to route queries to specialized agents: Smart Routing
• I want to execute tasks concurrently: Parallel Workflow
• I need to create agent workflows: Basic Workflow
• I want to discover agents automatically: Agent Discovery
• I need GitHub integration for my agent: GitHub MCP Provider
• I want browser automation capabilities: Browserbase MCP Provider
• I need secure file operations: Filesystem MCP Provider
• I need to add external tools to my agent: MCP Tools
• I want to build a complete application: Weather Assistant

Most examples can be run with:

python example_name.py

Some examples require API keys or additional setup. Check the comments at the top of each file for specific requirements and instructions.

# For basic examples
pip install python-a2a

# For server examples
pip install "python-a2a[server]"

# For LLM integration
pip install "python-a2a[openai]" "python-a2a[anthropic]" "python-a2a[bedrock]"

# For MCP (tool) support
pip install "python-a2a[mcp]"

# For everything
pip install "python-a2a[all]"

For a structured learning experience, we recommend following this sequence:

1. Start with the basics: Run through the Getting Started examples
2. Understand core concepts: Explore the Building Blocks examples
3. Add AI capabilities: Try the AI-Powered Agents examples
4. Implement streaming: Explore the Streaming examples for real-time responses
5. Learn agent networking: Experiment with the Agent Network examples
6. Extend with tools: Experiment with the MCP examples
7. Build complete applications: Study the Applications examples
8. Improve your workflow: Use the Developer Tools examples

Have an idea for a new example? We welcome contributions! Please check the CONTRIBUTING.md file for guidelines.

The python-a2a project includes a comprehensive validation system to ensure all examples work correctly. This is especially valuable before releases or after making significant changes.

# Validate all examples
python validate_all_examples.py

# Validate a specific category
python validate_all_examples.py --category streaming

# Skip slow examples for faster validation
python validate_all_examples.py --skip-slow

# Run examples concurrently for faster validation (be careful with server examples)
python validate_all_examples.py --concurrent 3

# Show detailed output for debugging
python validate_all_examples.py --verbose

The validation script provides detailed output including:

• Overall success/failure statistics
• Category-by-category results
• Detailed error messages for failing examples
• Skipped examples (due to missing dependencies or API keys)

Example output:

===============================================================================
VALIDATION SUMMARY
===============================================================================

Overall Results:
Total examples: 42
Passed: 39
Failed: 2
Skipped: 1

Category Results:
getting_started: [All 4 passed]
building_blocks: [All 4 passed]
streaming: [6/7 passed, 1 failed]
ai_powered_agents: [3/4 passed, 1 failed]
applications: [All 2 passed]
agent_network: [All 3 passed]
workflows: [All 3 passed]
mcp: [All 3 passed]
langchain: [All 4 passed]
developer_tools: [All 3 passed]

Failed Examples:
• streaming/06_distributed_streaming.py: Example timed out after 30 seconds
• ai_powered_agents/openai_agent.py: Missing API key environment variable: OPENAI_API_KEY

Skipped Examples:
• applications/openai_travel_planner.py: Skipped (missing API key: OPENAI_API_KEY)

--------------------------------------------------------------------------------
✗ Validation found 2 failing examples.
--------------------------------------------------------------------------------

• Run validation before submitting pull requests
• Run validation after making significant changes
• Include validation in CI/CD pipelines
• Use --category to focus on specific areas you're working on

All examples are released under the MIT License. See the LICENSE file for details.