MCP in production — CCA-F Exam Prep

L2.15|MCP in production

1/12

Countdown

A production dashboard. 3:47 PM on a Tuesday. An MCP server health monitor shows: Response time spiking from 200ms to 12 seconds. Active connections: 48. Error rate climbing: 2%, 5%, 12%, 23%. A Slack notification: 'Users reporting Claude lost access to tools.' A countdown timer in the corner: 'Time since last health check: 47 minutes.' Nobody noticed for 47 minutes.

3:47 PM. The MCP server starts dying. Nobody notices for 47 minutes.

A team deployed their MCP server to production. 50 users. HTTP transport. It worked great for two weeks. Then the database connection pool maxed out. Response times went from 200ms to 12 seconds. Timeouts started. Tools failed silently.

No health checks. No alerting. No monitoring. Users started reporting that Claude "couldn't access tools" but nobody connected it to the MCP server. By the time the team noticed, 47 minutes of conversations had been affected.

Building an MCP server is easy. Running one in production is where everything gets hard.