CrewAI (Agentic AI) + Snowflake MCP Integration

Simple project demonstrating how to integrate CrewAI agents with Snowflake via MCP (Model Context Protocol) to expose Snowflake resources and tools to AI agents.

---

🚀 TLDR - Just 2 Lines of Code

Add an MCP server to any CrewAI Agent definition:

support_agent = Agent(
    role="Customer Support Agent",
    goal="Analyze customer support cases",
    # ... other config ...
    mcps=[
        MCPServerHTTP(
            url=SNOWFLAKE_MCP_URL,
            headers={"Authorization": f"Bearer {SNOWFLAKE_PAT_TOKEN}"},
        )
    ],
)

That's it. Your agent now has access to all Snowflake MCP tools and resources.

---

Setup

1. Environment Variables

SNOWFLAKE_PAT_TOKEN=your_pat_token
SNOWFLAKE_MCP_URL=your_mcp_server_url
GOOGLE_API_KEY=your_gemini_key

MCP URL Format:

https://<org>-<account>.snowflakecomputing.com/api/v2/databases/<DATABASE>/schemas/<SCHEMA>/mcp-servers/<MCP_SERVER_NAME>

Example:

https://sfsenorthamerica-sg_demo13.snowflakecomputing.com/api/v2/databases/DASH_DB_SWT_2025/schemas/RETAIL/mcp-servers/mcp_server_swt_2025

2. Install & Run

uv sync
python main.py

How It Works

1. Snowflake MCP Server exposes tools (Cortex Search, Cortex Analyst, etc.)
2. CrewAI Agent connects via MCPServerHTTP
3. Agent automatically discovers available tools and uses them to complete tasks

Troubleshooting

SSL Certificate Issues? If you encounter certificate errors during testing, set verify=False in .venv/lib/python3.13/site-packages/mcp/shared/_httpx_utils.py (line 83). For testing only—not recommended for production.

I have logged issue here - crewAIInc/crewAI#4100

Resources

• Snowflake MCP Documentation
• CrewAI MCP Integration