Their data platform runs on Snowflake for warehousing, dbt for transformations, and Airflow for orchestration. When a pipeline breaks, the engineer opens Snowflake to check query history, Airflow to find the failed DAG run, and dbt Cloud to see which model failed. Three tabs, three auth flows, ten minutes of context-gathering before any actual debugging.
They wanted their AI agent (Claude in the terminal) to pull context from all three systems in one conversation.
Three MCP servers, each created in under a minute:
Snowflake — browsed the Snowflake web console, captured the query history and warehouse monitoring APIs. The agent can now check running queries, see warehouse credit usage, and pull recent query results without leaving the terminal.
dbt Cloud — browsed dbt Cloud, captured the run history and model status endpoints. The agent sees which models failed, reads the error logs, and checks lineage without opening a browser.
Airflow — browsed the Airflow web UI, captured DAG run status and task instance APIs. The agent checks which DAGs are running, which failed, and reads task logs inline.
When a Slack alert fires for a failed pipeline:
What used to take 15-20 minutes of tab-switching now takes one conversation. The engineer gets the full failure chain in seconds because the agent can access all three systems through typed MCP tool interfaces.
They tried building Python scripts for each tool. Each script took days to write, needed auth management, broke when APIs changed, and only did one thing. The DataFaucet approach captures whatever APIs the web UI already uses, handles auth, and deploys as a hosted endpoint their AI client connects to directly.
The three servers cost nothing on the free tier. Total setup time: about 4 minutes including browsing each app.
Pipeline incident response time dropped from 18 minutes average to 4 minutes. The agent handles the context-gathering that used to be pure human overhead. Engineers spend time fixing problems instead of finding them.
The team added a fourth server for their Grafana dashboards the same week. Same pattern: browse the UI, deploy the server, connect to Claude.
---
*DataFaucet turns any web application into an MCP server your AI agent can use. Create your first server free. Connect Snowflake, dbt, Airflow, or any tool with a web UI in 60 seconds.*
Related reading:
Create your Snowflake MCP server in 60 seconds.
Try with Snowflake →{
"mcpServers": {
"snowflake": {
"url": "https://datafaucet.dev/api/mcp/YOUR_SERVER_ID/sse"
}
}
}Replace YOUR_SERVER_ID with the ID from your DataFaucet dashboard after creating your Snowflake server.
Point DataFaucet at Snowflake and get a working server in 60 seconds.
Create Snowflake server free →After creating, add to Claude Desktop:
"snowflake": {
"url": "https://datafaucet.dev/api/mcp/YOUR_ID/sse"
}Free plan includes 3 servers. Upgrade to Pro for unlimited →
Top MCP servers for data engineers. Connect AI agents to Snowflake, dbt, Airflow, BigQuery, and pipeline tools via MCP.
Turn dbt Cloud into an MCP server. AI agents can check model status, view lineage, inspect test failures, and query run history.
Turn Apache Airflow into an MCP server. AI agents can check DAG runs, inspect task failures, and query pipeline metrics from Claude or Cursor.
See how DataFaucet compares
Point at any URL. Get a working MCP server in 60 seconds. No API docs needed.
Works with ChatGPT, Claude, Cursor, Copilot, Windsurf, JetBrains, and any MCP client
Get notified when new integrations launch
Join 500+ builders. New templates, guides, and MCP tips. No spam.