Lab 6: AI Agent + MCP

Difficulty: Advanced · Estimated time: ~5 hours

Objective

1. Build a Pydantic AI research agent
2. Create an MCP server with tools (web search, URL reader, SQLite storage)
3. Implement autonomous research loop
4. Generate structured reports
5. Deploy MCP server for VS Code integration

Step 1 — Create MCP server with FastMCP

python

from mcp.server.fastmcp import FastMCP
import sqlite3
import httpx

mcp = FastMCP("TDS Research Agent")

@mcp.tool()
async def web_search(query: str) -> str:
    """Search the web for information."""
    async with httpx.AsyncClient() as client:
        response = await client.get(
            "https://api.tavily.com/search",
            params={"query": query, "api_key": TAVILY_KEY}
        )
    results = response.json().get("results", [])
    return "\n".join(f"- {r['title']}: {r['url']}" for r in results[:5])

@mcp.tool()
async def read_url(url: str) -> str:
    """Read content from a URL."""
    async with httpx.AsyncClient() as client:
        response = await client.get(url)
    return response.text[:5000]

@mcp.tool()
def save_finding(topic: str, finding: str, source: str) -> str:
    """Save a research finding to the database."""
    conn = sqlite3.connect("research.db")
    conn.execute("INSERT INTO findings (topic, finding, source) VALUES (?,?,?)",
                 (topic, finding, source))
    conn.commit()
    conn.close()
    return "Saved!"

@mcp.tool()
def get_findings(topic: str) -> str:
    """Retrieve all findings for a topic."""
    conn = sqlite3.connect("research.db")
    rows = conn.execute("SELECT finding, source FROM findings WHERE topic=?",
                       (topic,)).fetchall()
    conn.close()
    return "\n".join(f"- {f} (source: {s})" for f, s in rows)

if __name__ == "__main__":
    mcp.run()

Step 2 — Build Pydantic AI agent

python

from pydantic_ai import Agent

research_agent = Agent(
    "openai:gpt-4",
    system_prompt="""You are a research assistant. Use tools to:
    1. Search the web for information
    2. Read URLs for details
    3. Save important findings
    4. Synthesize a report when done"""
)

result = await research_agent.run("Research the latest advances in RAG evaluation")
print(result.data)

Submission

GitHub repo with: MCP server, agent code, database schema, Dockerfile

Grading rubric

Criterion	Points
MCP server runs and exposes tools	25
Agent autonomously searches and reads	25
Findings stored in SQLite	15
Report generation works	20
MCP accessible from VS Code	15
Total	100