Explore — ask the analysis agent a question nobody wrote down

This agent reads and only reads: every analysis this plant computes, the state and tag history, the quality record, the audit trail and the trace beside this tab. It cannot book, propose, approve or change anything. Ask it something, and where the shape is the answer it draws the plant's own payload beside its words — the same chart the screens draw, stating the same total and the same coverage.

An exploration lives as long as this conversation does. It is not a dashboard and nothing here is published to anybody: a chart becomes part of a screen only when somebody decides it should be.

Conversations —

Every turn of every conversation this plant's AI has had: what was asked, which tools it called, what it proposed, what was done and what failed. The plant keeps its own copy, in its own database, so a conversation somebody had on a phone last night is readable here this morning and a bug is a trace rather than a screenshot.

Brains — what is on, and why the rest is not

What this screen promises

The trace records what the panel showed the person and one sentence per tool call. It never holds the system prompt, never the key, and never a tool's raw payload — a record of the plant's own rows read through this screen would be a way around the capabilities those rows are behind.

A plant in shadow mode writes these rows like any other: the trace changes nothing in the plant and carries nothing off the box. What shadow mode withholds is the cloud brain itself, so a shadow plant's rows are the local model's.

How far a person got through a walkthrough is not recorded. The walk runs in their browser and the plant is not told, and an unknown is written as an unknown rather than guessed at.

My agent

Everything this plant's analysis could say about you, shown to you. That is decision 0039 clause 4, and the test it sets for any analysis this product ever runs: if the plant would not show it to the person it is about, it should not be run. Nothing here is about anybody else, and there is no way to ask this screen about anybody else.

When an analysis named you —

Naming a person in an analysis needs the people.analyse capability, which no shipped role holds, and every answer that names one writes the row below. An empty list is a plant where no analysis has named anybody — which is a measurement, not a blank.

Your conversations —

Your questions, grouped —

The same rollup the analysis agent reads, narrowed to your own turns: which questions you asked more than once, and which you asked only once.

How long the trace is kept

Turns older than this are deleted as new ones are written, so what this screen shows is what the plant has. Zero keeps everything.

Changing it needs users.manage, which you do not hold.

The numbers that gate the assistant

These are Setup's, not this screen's: one Configuration entry per workspace, and a row per thing. Each opens where it is set.