I Connected DSPy to an MCP Server — Here’s What Happened

In this demo, I build and run a DSPy + MCP incident investigation agent. The project shows how to connect a DSPy ReAct agent to local MCP tools for investigating order and inventory issues, with optional distributed tracing using OpenTelemetry, Jaeger, and Logfire. GitHub repo: https://github.com/ekb-dev-ai/mcp-dsp... What’s included: DSPy ReAct agent MCP incident server Local order and inventory investigation tools Jaeger tracing with Docker Compose Logfire/OpenTelemetry setup Example incident workflow Run it locally: docker compose up -d python -m demos.incident_agent Setup: python -m venv .venv source .venv/bin/activate pip install -e ".[trace]" cp .env.example .env Chapters: 0:00 Intro 0:41 MCP server (context server) 05:33 Incident agent (DSPy class) 07:12 Demo implemetation 12:25 Run the demo and tracing UI check up This is a practical example of using MCP with DSPy to build agentic workflows that can call structured tools and trace what happens during execution. #DSPy #MCP #AIagents #Python #OpenTelemetry #Jaeger #AgenticAI