MCP Activity History

A searchable record of every tool call your agents made through 1DevTool — names, statuses, durations, errors, and safe previews.

When an agent does something surprising, the useful question is which tool did it call, with what, and what came back. Settings → MCP → Activity answers it.

What is recorded

Every MCP tool call your agents make through 1DevTool, kept across app restarts:

  • Tool name and status
  • Duration
  • Project and terminal it came from
  • The error, if it failed
  • A safe preview of the input and the output

Reading it

  • Filter by status to jump straight to failures
  • Expand a call for its full detail
  • Group by agent or by project to see who has been doing what
  • Copy the visible history when you want to paste it into a bug report
  • Load older entries to page back through the log
  • Clear retained activity when you want a clean slate

Redaction

Sensitive values are redacted and large binary results are shortened before they reach the interface. A screenshot does not arrive as a wall of base64, and a call carrying a token does not print it into a panel you might share.

The per-terminal view

Activity is the historical record; the live one is on the terminal itself. Each AI terminal shows badges for its recent tool calls — at most three at a time, with completed ones clearing after a short confirmation period. Click a badge for the call's details.

MCP badges and call details

When a call was refused

If an agent called a tool you have switched off in MCP Control, the refusal is recorded here too, and the terminal gets a badge naming the tool with a one-click Enable. That is usually the fastest explanation for "the agent said it couldn't do it".

Related