Find out what changed, when and by whom

Goal

List what was created or modified in a time window, see which authors were active, and correlate both with a baseline comparison. This is for a model steward, a reviewer, or an architect investigating an unexpected change. You end with a short, evidence-backed account of what was touched, when, by which authors, and what each change was.

Where it runs

This scenario needs the MCP server. Computer use covers its UI steps.

How it works

Two read-only operations answer complementary questions. get_updates_in_range returns every element, package or diagram whose modified date falls in the window. get_user_activity returns per-author counts for the same window. Both are in Audit trail. Neither says what changed on an item; for that, compare against a baseline from before the window (Protect a change with a baseline). Together the three answer "what changed, when, and by whom".

The author recorded on an item is the EA or Windows login stored on the element or diagram. Present the activity counts as the authors of the items modified in the window, not as proof of who made each individual change.

Before you start

• EA is running with the model open.
• A time window. A date alone is enough; a bare end date covers the whole day.
• Optionally, a baseline taken before the window, for field-level detail.
• An agreed way to handle the author names the report returns (see the privacy note below).

Steps

1. AI with computer use; otherwise you If EA is not running or the model is not open, start EA and open it. AI Power Tools ea-start-here runs the preflight. You fix the window and the scope.
2. AI Power Tools ea-change-management lists what was touched in the window. The kind can be elements (the default, which includes packages stored as objects), packages, diagrams, or all.
3. AI Power Tools ea-change-management scopes the result to the package of interest by filtering the list or by querying the model directly.
4. AI Power Tools ea-change-management reports which authors were active.
5. AI Power Tools ea-change-management gets field-level detail by comparing against a pre-window baseline, or by reading individual elements.
6. AI ea-change-management correlates the results and reports. You decide what to do about it.

Step 2, the window. Give start and end as a date, a date and time, or an ISO-8601 timestamp. EA stores local times with no timezone, so a timezone designator is ignored and the response says so; the wall-clock value is used as given. The response shows the window as it understood it next to the one you asked for, then lists each item with its kind, name, type, package and modified time. An unreadable date or an unknown kind returns an error rather than an empty list. Connector-level history is not tracked.

Step 3, scope. Neither operation takes a package. Both are repository-wide, so the assistant filters the result by package afterward.

Step 4, privacy. The result is one row per author with the number of elements and diagrams modified and when each was last active. The author value is the raw EA or Windows login. Handle it as a personal identifier: fine to read inside your own repository for audit, not for pasting into shared tickets or documentation.

Step 5, detail. The audit operations show that an item changed, not how. For how, compare against a baseline from before the window. "Changed" items show the property values before and after, "Model only" items are additions, and "Baseline only" items are deletions. Deletions leave no row in the audit operations, so a baseline is the only way to see them.

Prompts to use

Opening:

Follow-ups:

• Detail: "For the five elements changed after noon, compare against the baseline from the start of the day and show me what each change actually was."
• Scope: "Only show me the part inside the Payments package."
• Deletions: "Did anything get deleted? The change list wouldn't show that."
• Hand-off: "Summarize it for the review: counts only, no names of people."

Recommended models

Ratings are guidance based on the kind of reasoning the scenario needs, not measured benchmarks.

Checking the result

• Correct. The window the assistant reports matches what you intended (check for a note that a timezone was ignored). Spot-check two items: the modified time in the report equals the value in EA's element properties. Red flag: a window that differs from the one you asked for with no explanation.
• Complete. Compare the item count with a direct count over the same window for the kind you asked for. Remember the operations cover elements, packages and diagrams only; connectors and deletions are not in them. Red flag: a report that claims no deletions without a baseline comparison to back it.
• Nothing unintended changed. These operations are read-only. If you also compared against a baseline, that opened EA's comparison view but did not change the model.

Variations

• Between two baselines. Baseline the model root on a cadence you arrange, and use the two baseline dates as the window.
• Before a restore (Protect a change with a baseline): check the window since the baseline to see whose work a restore would discard.
• Focus a hygiene pass on what changed recently (Clean up model decay).

Limitations

• Both operations are repository-wide; the assistant filters the result by package afterward.
• No connector history, no deletions. Use a baseline comparison for deletions.
• A modified date can show that an item was saved, not that its content differs. A busy window can contain many touched items. Use the baseline comparison to separate real changes from saves.
• Nothing marks how a change was made. The output does not say which changes an AI session made and which were made in EA's UI.
• Author names are raw logins. Do not copy them into shared material.
• No timezone conversion. The wall-clock window you give is compared to EA's stored local times.