The screens an agent product needs
31 screens, from the chat to the audit log. Each has two or three ways to lay it out and a wireframe of each, the Candor items that build it, an install command, and a brief you can hand to a coding agent. Where a pattern already covers part of a screen, it links there.
Talking to the agent
Where a person and an agent work through a request together: the chat, the ask before acting, the plan, the copilot and the handoff to a person.
- Agent chatTalk to an agent, watch it work, and see what it did and what it used.3 layouts
- Permission prompt in chatLet the person allow or deny one tool call before it runs.3 layouts
- Plan reviewShow what the agent is about to do as steps a person can trim, question and accept before anything runs.3 layouts
- Support copilotDraft answers for a support agent from the help docs, with sources, for a person to approve, edit or reject.2 layouts
- Human handoffPass a thread from the agent to a person, and between people, without losing where it stands.3 layouts
Approving and controlling
Where people decide what an agent may do and undo what it did: approvals, tool permissions, budgets and restore points.
- Approvals inboxOne place where people decide what agents asked to do, each decision tied to the exact change.3 layouts
- Environment promotionMove a configuration from one environment to the next with the diff, who approved it and what can be rolled back.2 layouts
- Tool permissionsSay, per tool, what an agent may do without asking, what needs a go-ahead, and what it may never do.3 layouts
- Cost guardSay what a run is likely to cost before it starts, hold it to a budget, and report what it did cost.3 layouts
- Checkpoints and restoreLet a person go back to an earlier point in a run, knowing what going back does and doesn't undo.3 layouts
Watching the work
Where people see what a run did, what it cost and what happened to it: run detail, traces, the console and the audit log.
- Run detailShow one run: what it was asked, what it did, how it ended and what it cost.3 layouts
- Trace viewerShow what happened in a run in order: which step called which tool, how long each took, and where it broke.3 layouts
- Run console with recoveryList every run, steer the ones in flight, and say what to do about the ones that stopped.3 layouts
- Audit logRecord what the agent and the people did, searchable, exportable, and with retention that says what is kept.3 layouts
- Reports and dashboardsShow how the work is going in charts and tables, and hand a finished report to someone who wasn't there.3 layouts
Judging the output
Where people decide whether the agent is good enough: eval scorecards and the queue for grading real answers.
- Eval scorecardShow how an agent version did on a fixed suite, against the baseline, with what to look at first.3 layouts
- Eval review queueLet people grade real agent answers one at a time, fast, and turn the failures into work.3 layouts
- Document reviewRead a document or file an agent produced beside the work, compare it with what was there, and decide what to keep.3 layouts
- Extraction reviewCheck the values an agent extracted from a document against the page, correct what's wrong and send what the person accepted.3 layouts
Knowledge and setup
Where the agent is configured and taught: its settings, what it knows, the servers it connects to, and who may change any of it.
- Agent roster and settingsSee every agent on the team, and change one agent's model, instructions, tools and schedule, with a history of changes.3 layouts
- Knowledge base and retrievalShow what the agent knows, where it came from and how fresh it is, and let people see the passages behind an answer.3 layouts
- Connecting an MCP serverLet a person decide whether an agent may use an MCP server: where it runs, who runs it and what it can do.3 layouts
- Admin: roles and data controlsSay who can do what, bring people in, and show what happens to the account's data.3 layouts
- Account and personal settingsLet a person change who they are in the app, how it reaches them and where they are signed in.2 layouts
- Import dataTake a spreadsheet in, match its columns to the product's fields and check the rows before anything lands.2 layouts
The app around the agent
The frame that holds the rest: the shell and its session list, project workspaces, and what reaches people while they're away.
- Session list and app shellHold the product together: where you are, your threads, the main places and the search.3 layouts
- Home dashboardShow a person what needs them and what happened since they were last here.3 layouts
- Project workspaceGive each project a conversation with an agent team, and show which projects need the person.2 layouts
- Notifications and the away digestTell people what happened while they were away, in the order they'd act on it, and let them choose what reaches them.3 layouts
- Sign-in and first runGet a person signed in, into the right account and through the setup that comes before the first useful answer.3 layouts
- Billing and usageShow where the subscription stands, what has been used against each limit, and what a change of plan would take away.3 layouts