Dashboard Overview
The screen an app opens on: shell, header, KPI row, chart and activity feed — numbers first, shape second, changes last.
The three states an agent screen usually omits: the wait before the first token, the trace after something broke with retries costed in the open, and the diff it wants a person to accept.
app/agent-inspector-page.tsxnpx hoverlab add agent-inspector-page
Or over MCP, from your editor's agent — no account needed.
Free to read, copy and install, for personal and non-commercial projects. Shipping it in client work or a paid product needs Pro ($79 once). The source lands in your repo and stops being ours — no attribution, nothing to upgrade.
What is running, what it looks like while you wait, what it looks like when it breaks, and what it is asking you to approve.
2 of 5 complete · 1 failed
Complete: Pull cancellations for Q3
2,481 rows from warehouse.subscriptions
Complete: Join to account seat counts
Matched 2,463 of 2,481 — 18 accounts deleted
Failed: Fetch exit-survey responses
CRM returned 429 — rate limited after 40 of 300
Running: Score accounts by churn risk
Running the model over the joined cohort
Queued: Draft the summary
Reading your question
Paused on step 3 of 5 · 211s · $0.14 spent so far
Read the deploy manifest — Done
infra/deploy.yaml, 240 lines
Resolve the image tag — Recovered
Registry timed out; used the digest pinned in the lockfile instead.
Recovered — what the later steps used came from the fallback, not from the registry.
Apply the migration — Failed
Blocked. The migration wants a lock the reporting job is holding.
Restart the workers — Waiting
Not started — blocked by the step above.
Verify the health check — Waiting
Not started.
What the tool returned
ERROR: could not obtain lock on relation "orders" (SQLSTATE 55P03)
detail: process 4417 holds AccessExclusiveLock
hint: retrying will queue behind the same lockWhat the agent thinks that means
The nightly reporting job took an exclusive lock on orders at 02:00 and has not released it. Retrying as-is will queue behind the same lock and time out again. Either wait for that job, or run the migration with a lock timeout and accept a partial apply.
I filled gaps and corrected classifications from public sources. Row 4 is a guess from billing inactivity; the rest are verified.
| Record | Field | Change | |
|---|---|---|---|
| Northwind Retail | Industry | ||
| Vela Logistics | Employees | ||
| Harbour Foods | Owner | ||
| Pike & Sons | Status | ||
| Cobalt Interiors | Industry |
5 of 5 will be written
Suggested action
Demand is tracking 22% above last October and your supplier holds a volume break at 1,200 units. Ordering today lands stock four days before the forecast stockout.
Based on 3 seasons of sales history. The forecast has not seen a promotion like the one running this month, which is why this is not higher.
The real page, rendered in your current theme — every section below is a live block, not a screenshot.
Only want one section? Open it and copy that block instead — browse Agent Reasoning.
/**
* What the agent is doing, what it did, and what it wants you to accept.
*
* `agent-run-page` next door is the happy path: a trace, in order, ending
* in an answer. This is the screen for the other three states, and they are
* the states that decide whether anyone trusts the thing.
*
* The order is by how much of the user's attention each one is entitled to.
*
* task list the run, as rows. Ambient. Glanced at.
* working the gap before the first token, which is where people
* decide the product is broken
* failure the trace after something broke, with the retries and
* what each one cost counted in the open
* diff review the machine's proposed edits, accepted one at a time
* confidence how sure it was, as a native meter and in words
*
* <AgentWorkingIndicator> is second rather than buried because an empty
* pane during a thirty-second tool call is indistinguishable from a hang,
* and "indistinguishable from a hang" is a support ticket. It is the
* cheapest component on this page and the one that changes the most.
*
* The diff review is deliberately after the failure trace. Reviewing
* proposed changes is the moment a person is asked to take responsibility
* for the machine's work, and they should reach it having already seen what
* this system looks like when it is wrong.
*
* Confidence is last for the same reason a footnote is last: it qualifies
* everything above it, and a score at the top of a screen gets read as a
* grade rather than a caveat.
*/
import * as React from 'react'
import { AgentTaskList } from '@/components/agent-task-list'
import { AgentWorkingIndicator } from '@/components/agent-working-indicator'
import { AgentRunFailure } from '@/components/agent-run-failure'
import { AgentDiffReview } from '@/components/agent-diff-review'
import { ConfidenceRecommendation } from '@/components/confidence-recommendation'
export default function AgentInspectorPage() {
return (
<main className="min-h-screen bg-background text-foreground">
<section className="mx-auto w-full max-w-5xl px-6 pb-2 pt-12">
<h1 className="text-2xl font-bold tracking-tight">Run inspector</h1>
<p className="mt-1 text-sm text-muted-foreground">
What is running, what it looks like while you wait, what it looks
like when it breaks, and what it is asking you to approve.
</p>
</section>
<AgentTaskList />
<AgentWorkingIndicator />
<AgentRunFailure />
<AgentDiffReview />
{/* A caveat, so it reads as one. */}
<ConfidenceRecommendation />
</main>
)
}
components/ — each block page has its own copy button.app/agent-inspector-page.tsx. The imports already point at @/components/…, so they resolve with no edits.The component, its props, the design tokens it expects and the command that installs it — as one prompt. Paste it into Claude, Cursor, v0 or ChatGPT and what they build around it will match the rest of the catalog instead of inventing its own system.
The screen an app opens on: shell, header, KPI row, chart and activity feed — numbers first, shape second, changes last.
An agent working inside a real app: a transcript down the middle — reasoning, answer, then the one card that can act — with what it may read and what it noticed demoted to a rail.
A kanban board inside the app shell, with tabs that admit the board is one view among several and a header whose primary action is creating work.
The canonical CRUD list — header, toolbar, sortable table and pagination framed as one panel instead of three stacked cards.
A wizard for the session someone finishes and a checklist for the one they do not, with the only genuinely blocking step in front of both.
Built around the state a search page is in most of the time — empty — so recent queries lead and the facets sit in a sidebar that survives growing to twelve.
One run opened up in the order you would debug it — intent, actions, the retries that are invisible in both, then cost.
The queue, the same queue sorted by how long things have waited, and the policy that decided both — with the policy last so it is not scrolled past forever.
The blank thread answered with real starter prompts, the model picker with the three numbers the choice turns on, and the four surfaces a session grows into — attachments, branches, canvas, threads.
Why the answer said what it said: what is connected, how stale it is, what the query was allowed to see, what fitted in the window and what fell out — plus the failed-search state a debugging tool is most needed in.
Five ways to reach a model without the cursor leaving the sentence — slash menu, inline completion, selection toolbar, action menu, property inspector — and the two surfaces that say where the answer came from.
A table after real people have used it — columns reordered from the keyboard, a selection bar that asks page-or-all, subtotals that survive collapsing, cells edited in place, and the empty state its own filters produced.
The saved view and the comparison window first, because every figure below depends on both — then six charts, none of which load a charting library.
Thresholds with how often each has actually fired, what that produced, where it went — and the digest settings that exist because of the answer.
The flow between an empty product and a useful one, in the order it goes wrong: sample data offered with equal weight, a file, a URL, the column mapping everything stalls on, and an upload that survives the connection dropping.
A profile form where the role options state what they actually change, the avatar crop works from the keyboard, and a navigation guard names the fields it is about to throw away.
The furniture that is on every screen and belongs to none — navbar, mobile drawer, command palette, shortcut sheet, header search, scope switcher, and the two shapes a detail view takes without losing the list.
Backwards from now and forwards from it on one screen — a day-grouped timeline, a month grid where a busy Tuesday never changes the row height, and the state both are in on day one.
What the product looks like while onboarding happens: a tour step anchored to the control it explains, the skeletons of the slowest load it will ever do, and the checklist that outlives both.