Documents
Welcome
Introducing
Why Misaki Studio
Compare
Start Here
Learn
Workflows
Export & Ship
AI & MCP
What MCP means in Misaki Studio
Connect an agent
Select a project safely
Target files, folders, and elements
Build with prompts
Combine UI and agent work
Review agent activity
Verify agent work
Undo agent changes
UI-to-MCP capability map
Example prompts
MCP limitations
Questions & Answers
Reference
Troubleshooting
Examples

Review agent activity

The Misaki Studio desktop editor can show where MCP work is happening across file tabs, the File Explorer, responsive screen frames, and the Element Explorer. Use these marks to find an agent's work, then inspect and verify the result itself.

Coverage: Desktop UI review of MCP activity. Activity marks are editor-only view state retained for the current session until explicitly cleared; they are not stored project content or a permanent audit log.

Where activity appears

When a project is selected and editor activity tracking is available, a tool call is bracketed as it starts and finishes. Affected targets can appear on:

  • an open file tab;

  • a file or folder row in the File Explorer;

  • a responsive screen frame on the canvas;

  • an element or one of its ancestors in the Element Explorer.

If an affected file, folder, or element is hidden inside a collapsed branch, the editor rolls activity up to the deepest ancestor that is currently visible. Expand that branch to locate the more specific target.

What a mark means

Activity moves through writing, settled, and unseen states. The exact visible treatment can change with the interface theme, so interpret the state and tooltip rather than relying only on one color.

A mark means the editor attributed recent MCP activity to that target. Reads and writes both produce marks when the call identifies a file or node; the tooltip arrow distinguishes whether project rows actually changed. A mark does not mean:

  • the requested outcome is correct;

  • the call changed every value it intended to change;

  • generated output builds;

  • a publish or external write succeeded;

  • no other file was affected.

The read/write indicator is derived from whether the operation actually wrote rows, rather than from a guess based only on the tool name. A write-oriented tool that changed nothing therefore appears as a read.

Read the activity history

Hover the activity indicator to inspect recent calls. The history is newest first and can show:

  • the operation title and target;

  • whether it read or wrote;

  • whether it is still running;

  • whether it failed.

The tooltip shows the first five matching entries and summarizes additional older entries. The underlying in-memory history is a fixed recent ring rather than an unlimited log.

Use this history to answer “what should I inspect next?” A failed entry or an unexpected target is a reason to stop and read current state before retrying.

Clear a mark

Hover the indicator and choose Clear to dismiss it. Time passing, opening or focusing a file, and selecting an element do not clear activity.

Clearing a file indicator dismisses that file mark wherever it appears. Clearing a rolled-up folder, screen, or element-container indicator dismisses the currently represented descendant marks while leaving unrelated activity alone. Dismissing a mark does not erase the bounded recent activity history or change project content.

Clearing is a visual acknowledgement only; it is not confirmation that the result passed verification.

A practical review sequence

  1. Wait for the active write indicator to settle.

  2. Hover the mark and confirm the operation, target, write status, and failure state.

  3. Expand rolled-up Explorer branches until you find the affected file or element.

  4. Read the current values or hierarchy.

  5. Preview or interact with the result when behavior or rendering changed.

  6. Inspect generated output or the external destination when the task crossed the project boundary.

  7. Only then choose Clear or continue editing.

Attribution limits

Activity is most precise when an operation reports a file UUID, node ID, or parent node ID. Calls without a direct file target can sometimes be attributed by observing which project row changed. That makes the feature useful for orientation, but not suitable as a compliance or audit system.

Other limits:

  • activity is available in the desktop/editor environment, not as a durable project record;

  • only a bounded recent history is retained;

  • clearing UI marks does not delete or undo project content;

  • reads and writes can both leave an activity mark when the call reports a target;

  • a correctly attributed write can still produce a wrong design.

Screenshot status

A current Website-project screenshot showing live and settled activity marks has not yet been captured for this page. The screenshot remains unfinished until the documented states are exercised without changing unrelated project content.

On this page
Review agent activity
Where activity appears
What a mark means
Read the activity history
Clear a mark
A practical review sequence
Attribution limits
Screenshot status