@deskdaemon: Save this for when you set up your first one. These MCPs can touch more than you think. I counted them from my own session logs. Here's what each one can reach. 👇 READ BELOW What it is: An MCP is a plug that lets Claude Code use another app or service for you, like your browser or a database. These four are the ones I actually have connected, and the call counts come from my own session logs. How to use it: 1. Pick one job you want Claude to do, like looking up current docs or editing a database. 2. Connect the matching MCP, then read the permission screen before you click allow. 3. Give it the smallest account that works, like a dev project instead of production. 4. Watch what it does for a few sessions before you trust it with more. Why it matters: You get Claude doing real work in your tools instead of you copying and pasting between them. Knowing what each plug can touch means you stay in charge of it. Why not: An MCP can do anything its account can do, so skip it, or use a throwaway account, if the account holds something you can't afford to lose. In my setup: Every MCP on this list is connected to my Claude Code, and the call counts come from my own session logs. Safety: license each MCP's own · last commit varies · connectors sign in with OAuth: read the consent screen, local servers run code on your Mac: read the repo first. Save this for the next time you're about to click allow on an MCP. #claudecode #aitools #devtools #buildinpublic #mcp

There are no more comments for this video.

To see more videos from user @deskdaemon, please go to the Tikwm homepage.