Frequently Asked Questions (FAQ)

Q: What is the A2A MCP Server? A: The A2A MCP Server bridges the Model Context Protocol (MCP) with the Agent-to-Agent (A2A) protocol, enabling MCP-compatible AI assistants (like Claude) to interact with A2A agents.

Q: What are the key features of the A2A MCP Server? A: Key features include agent management (register, list, unregister), communication (send messages, stream responses), task management (track, retrieve, cancel), and transport support (stdio, streamable-http, SSE).

Q: What is MCP? A: MCP (Model Context Protocol) is an open protocol that standardizes how applications provide context to LLMs.

Q: What is A2A? A: A2A (Agent-to-Agent Protocol) enables communication and interoperability between different AI agents through a standardized JSON-RPC interface.

Q: How do I install the A2A MCP Server? A: You can install it via Smithery or pip (pip install a2a-mcp-server) or by cloning the repository from GitHub and installing the dependencies.

Q: How do I configure the A2A MCP Server? A: Configuration is done using environment variables such as MCP_TRANSPORT, MCP_HOST, and MCP_PORT.

Q: What transport types are supported? A: The server supports stdio (default), streamable-http, and SSE transport types.

Q: How do I configure the server for Claude Desktop? A: You need to modify the claude_desktop_config.json file and set MCP_TRANSPORT to stdio. Use the provided config_creator.py script to automate this process.

Q: How do I use the A2A MCP Server with Claude Web? A: Start the MCP server with the streamable-http transport, enable the MCP URL connection in Claude’s Tools menu, and use the server’s URL (e.g., http://127.0.0.1:8000/mcp).

Q: What MCP tools are available? A: Available tools include register_agent, list_agents, unregister_agent, send_message, send_message_stream, get_task_result, and cancel_task.

Q: How do I register an A2A agent? A: Use the register_agent tool with the agent’s URL.

Q: How do I send a message to an agent? A: Use the send_message tool with the agent’s URL and the message.

Q: How do I retrieve a task result? A: Use the get_task_result tool with the task ID.

Q: What do I do if an agent can’t be registered? A: Verify the agent URL is correct and accessible and that the agent has a proper agent card at /.well-known/agent.json.

Q: What do I do if messages aren’t being delivered? A: Ensure the agent is registered (use list_agents) and that the agent is running and accessible.

Q: Where can I find the claude_desktop_config.json file? A: - Windows: %APPDATA%Claudeclaude_desktop_config.json

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

Q: What is UBOS? A: UBOS is a Full-stack AI Agent Development Platform that helps you orchestrate AI Agents, connect them with your enterprise data, build custom AI Agents with your LLM model and Multi-Agent Systems.