While a Mind works on an answer, every step it takes appears in the chat as a card: a database it reads, a query it runs, a web page it opens, a thought it works through. This page explains the cards and the Tool History panel that collects them. The cards of a specific area are described in Warehouse tool cards in chat, Workflows in chat and Knowledge search and skills in chat.
12 rows, 1 row or 0 rows. When a limit cut the result, the card reads 40 out of 250 rows; when it shows one page of a larger result, 1 - 50 out of 900 rows.{name} v. when the query has a version.Showing 1-50 out of 250 rows; it appears on a desktop in Cached mode. Export appears once the result has more than 50 rows; see Export and copy.On a desktop, point at a card and select Maximize; the step opens in a large centred view. On a phone, tap the card itself, and the view opens full screen. What Maximize opens depends on the step:
| Step | What opens |
|---|---|
| Query steps | The query view: SQL and results |
| List Databases, List Tables, Describe Table | The table view of the step |
| Web Search, Read Webpage and knowledge search steps | The results view of the step |
| A slide deck or presentation step | The artifact itself, once it exists; before that, the step in Tool History |
| Swarm steps and benchmark answer steps | The step's detail in Tool History |
| Other steps in a Warehouse Builder or Lake Builder chat | The step's detail in Tool History |
Open details on a card always opens the step in the Tool History panel, and a chip on a card opens its step too. The artifact view that a deck step opens is described in Artifacts.
7 results. In its full view each result shows its position, title and link; links open in a new tab, and Copy Results copies the list. While it works, the view reads Initializing... or AI is searching the web...; if a response is not in a shape the view can read, the raw text shows instead.3/, revisions and branches marked as it goes, and a planning step that reads Reflected and planned with What I Learned, My Reasoning and Next Step. Thinking History collects the sequence. These steps fold on their own when the answer finishes.3/. A finished plan folds itself and reads 7/.Tool History is the panel that lists the steps behind an answer. Open it from the chat header, or press Ctrl+Shift+H; the panel toggles are described in The Minds workspace.
A step that arrives in a shape the chat cannot draw is replaced by a small notice that reads Couldn't display this tool result, with a Try again button. While an answer is still streaming, a broken step draws nothing and the chat keeps working; the notice stays only if the step is still broken once streaming ends. Try again redraws the step, and the notice clears by itself when the step's content changes.
Success means the step finished. Error means it failed; open the step to read why. Awaiting approval means it is held until you approve or reject it. Rejected means a held step was rejected, Skipped means it did not run, and Pending means it has not started yet. Approvals are described in Change approval in a builder chat and Stop and abort runs in chat.
On a desktop, point at the card and select Maximize for the large view, or Open details to open the step in Tool History. On a phone, tap the card. Clicking the card body itself only folds or unfolds its brief details.
Each step opens the view that fits it: a query step opens its query view, a catalogue step opens its table, and a web or knowledge step opens its results. A slide deck or presentation step opens the artifact itself once it is built. Swarm steps and benchmark answer steps have no large view, so they open in Tool History instead; in a Warehouse Builder or Lake Builder chat, steps outside the query family open there too. See Artifacts for the artifact view and Swarm agents for swarm steps.
A limit cut the result: the step returned 250 rows and the view shows 40 of them. A form like 1 - 50 out of 900 rows means the view shows one page of a larger result, and you can move to another page. 0 rows means the query came back empty; open the step to read the SQL it ran.
Steps that run a stored query carry only its name. The SQL block is filled in from the step that created or last changed that query in the chat, so it can appear on a step where you never wrote it. If that step is not in the chat, the SQL block does not show; look for the query in Artifacts, where saved queries keep their SQL.
They show the Mind's reasoning while it works: a step-by-step reasoning step tracks progress such as 3/ and marks revisions and branches, and a planning step reports what it learned and what comes next. Thinking History collects the sequence. These steps fold on their own when the answer finishes.
Three kinds of step do not appear: the Mind's step-by-step reasoning, deck and presentation generation, and the questions the Mind asks. Open the artifact for a deck or presentation step, and answer questions from their panel (Questions from a Mind). Every other step, including the Mind's own thinking records, is listed.
It is the Mind's plan for the current work: 3 of 7 steps are done. Open it to see each step and its state (Completed, In progress, Pending or Skipped). The Mind updates the plan as it works, and a finished plan folds itself and reads 7/.
Type in Search in tools...; it matches step names, statuses and the content of steps. Use Filters to show one kind of step, All Tools to clear that pick, and Reset to clear the search and the filter together. If nothing matches, the list reads No tools match your filters.
You asked to open a step in full: from its card, through Open details, or through a chip on the card. The panel opens to show that step in its place. Select Go to tool history to return to the list, or close the panel.
That step's content could not be drawn. Select Try again; if it stays broken, reload the page and open the step once more, and ask the Mind again if you still need the result. While an answer is still streaming, a step like this draws nothing and the chat keeps working.