Protocol Bridge: MCP <-> A2A

The missing infrastructure layer between Model Context Protocol (agent-to-tools) and Agent-to-Agent Protocol (agent-to-agent). A bridge that lets Claude Code agents collaborate with Gemini, GPT, and other AI systems via standardized protocols.

What It Does

• MCP to A2A: Expose MCP tools as A2A Agent Cards
• A2A to MCP: Allow A2A clients to call MCP tools
• Multi-vendor: Enable Claude + Gemini + GPT agent collaboration

Installation

pip install protocol-bridge

Quick Start

from protocol_bridge import Bridge

bridge = Bridge()

# Expose MCP tools as A2A Agent Cards
bridge.expose_as_a2a(mcp_server="my-tools")

# Call MCP tools via A2A
result = bridge.a2a_to_mcp(
    agent_card="gemini-agent",
    tool="code_search",
    params={"query": "authentication"}
)

Demo

Claude Code agent delegates a task to a Gemini agent via A2A, which uses MCP tools to complete it:

python -m protocol_bridge.demo

Why This Exists

MCP won the tool-access war. Google's A2A won agent communication. But the two protocols don't talk to each other. 50+ companies support A2A. Thousands of MCP servers exist. No bridge. This is the middleware layer for the entire multi-agent ecosystem.

Status

Early development. See BRIEF.md for project plan.

License

MIT