One Claude Code Tip a Day: Audit Tool Connections With /mcp
Use /mcp as a Claude Code preflight: audit connected tools, name what the session can prove, and turn missing access into concrete verification checks.
This post is part of the “One Claude Code Tip a Day” series — a daily guide to using Claude Code more effectively.
The worst time to discover a missing tool connection is after Claude Code has already designed the fix around it. I have seen sessions waste twenty minutes planning a GitHub triage around an unavailable MCP server, or suggesting database checks when no database tool was connected. The patch was not bad because Claude was lazy. It was bad because the session started with a false assumption about its tools.
Today’s habit is to run a short tool audit with `/mcp` before a task depends on external context. Official Claude Code docs describe `/mcp` as the command for listing connected MCP servers. In practice, treat it as a preflight checklist: what can this session really see, and what must be verified another way?
The real failure: planning around a tool you do not have
Imagine debugging a production-only checkout failure: “Stripe webhook succeeds, but paid orders remain pending.” A confident prompt might ask Claude to check Stripe logs, compare payloads, inspect deploys, and patch the order state transition. But if your session has no Stripe MCP server, no database connection, and no deploy logs, that plan quietly turns into theater. Claude may infer what the logs probably say and write code against a guessed failure mode.