MCP Search

MCP Client with A2A Integration Overview This Python-based MCP (Model Control Protocol) client provides a robust interface for interacting with MCP services while incorporating A2A (Agent-to-Agent) protocol support and Weave observability.

Features MCP Protocol Support Full MCP 2024-11-05 protocol implementation Session management and initialization Tool discovery and search capabilities Detailed tool information retrieval Tool configuration management A2A Integration Dynamic agent discovery (ports 1000-1010) Agent card fetching from well-known endpoints LLM-powered agent routing Conversation history management Multi-agent task coordination Weave Observability Automatic MCP call tracing Custom operation spans Performance metrics Error tracking Agent interaction visualization Configuration Environment Variables GEMINI_API_KEY: Required for LLM-based agent routing ANTHROPIC_API_KEY: Optional for Claude-based tools WEAVE_PROJECT: Defaults to "wv_mcp" MCP Settings Protocol Version: 2024-11-05 Default Service URL: https://mcpsearchtool.com/mcp Default Transport: http-only A2A Configuration Host: localhost Port Range: 1000-1010 Agent Card Path: /.well-known/agent.json Default Model: openai/gpt-4o-mini Usage Basic MCP Operations Initialize Client:

Create MCPClient instance Initialize connection Verify session establishment Tool Management:

List available tools Search tools by natural language Get detailed tool information Add tools to configuration A2A Integration:

Discover available agents Route requests to appropriate agents Handle agent responses Maintain conversation context Advanced Features Media Handling:

Text processing Image support (PNG, JPEG) File attachments Binary data handling Error Management:

Comprehensive error tracking Automatic retries Fallback mechanisms Error reporting via Weave Observability:

Request tracing Performance monitoring Error tracking Agent interaction analysis Error Handling Common Errors Connection failures Authentication issues Invalid tool configurations Agent discovery problems Routing failures Recovery Strategies Automatic session renewal Connection retries Agent failover Graceful degradation User feedback

Built With

• a2a
• cloudflare
• crewai
• mcp
• python
• rust
• vector
• weave