Agent run trace

AI token usage panel with a monthly progress bar, a per feature breakdown, and the overage policy stated in the same view.

Conversation thread with alternating user and assistant bubbles, inline source chips, and regenerate and copy affordances.

Pre-conversation assistant panel with 4 example prompt chips and plain lists of what the assistant can see and cannot do yet.

Sources side panel listing 4 cited documents with title, matched snippet, relevance chip, and an honest footer note.

Data access panel listing assistant readable sources with icon, item count, last sync time, and a switch, plus one excluded source.

Chat history sidebar listing past assistant threads grouped by recency with title, snippet, timestamp, and a delete affordance.

AI response feedback row with thumbs up and down, a copy and regenerate action, and a reason picker revealed on negative feedback.

Ruled rows pairing each safety rule with an icon, explanation, and switch, above a blocked topics list and an audit logging note.

Text editor surface showing an AI ghost-text completion with Tab to accept and Esc to dismiss keyboard hints and an authorship notice.

A header switch and ruled list of remembered facts, each with its source, date, and a forget button, above a permanent deletion note.

Model selection card with three rows showing tier, cost per 1k tokens, a one-line tradeoff, and the balanced model pre-selected.

AI input surface with a model chip, attach and slash command buttons, a token counter, and a keyboard hinted send button.