All blogs
How-to

Start and stop the Memoir UI from Claude Code

No flags, no ports to remember. Just say "start memoir ui" and "stop memoir ui" — Claude Code launches and tears down the browser-based memory inspector for you.

Starting and stopping the Memoir UI inside a Claude Code session.

Two phrases, full lifecycle

The Memoir UI is a browser-based inspector for your memory store — branches, commits, blame, and the live taxonomy tree. It's how you look at the memories your agent has been writing without leaving the terminal flow.

Inside Claude Code, you don't need to remember a CLI invocation or a port number. Two natural-language phrases drive the whole thing:

  • "start memoir ui" — Claude Code launches the UI server and opens it in your default browser. If you prefer an explicit invocation, the slash command /memoir:ui does the same thing.
  • "stop memoir ui" — Claude Code tears the server down. Then close the browser tab and you're done.

What happens under the hood

When you say "start memoir ui" — or run /memoir:ui directly — Claude Code launches the Memoir UI server in the background, waits for it to become healthy, and opens the inspector in a new browser tab pointed at your current store.

When you say "stop memoir ui", Claude Code stops the background process so the port is freed for next time. The browser tab itself is harmless once the server is down — close it whenever you like.

When to use it

  • After auto-capture has been firing for a while — quickly scan what got written and on which branch.
  • Before a merge — eyeball the pending memories on your feature branch instead of guessing from CLI output.
  • Debugging a bad recall — use blame to trace a value back to the session that wrote it.

The whole loop is designed to be cheap: open it, look at the thing, close it, get back to coding.

Get Memoir running