Skip to content
pandacoderz/ui

Agent Run

Live agent plan with nested tool calls, a streaming log, and a human approval gate.

Open in new tab ↗

What’s inside

  • Steps shows the plan with pending, running, completed, failed, and waiting states.
  • Tool cards nest inside each step’s collapsible details.
  • Reasoning shows the agent’s rationale on steps that have one.
  • Approval gate pauses the run until you approve or deny, then continues or stops.

Installation

npx shadcn@latest add https://ui.spencerwueste.com/r/agent-run.json

The registry item pulls in every component it depends on.

Usage

import AgentRun from "@/components/blocks/agent-run/agent-run";

export default function Page() {
  return (
    <div className="h-dvh p-4">
      <AgentRun />
    </div>
  );
}

Give the parent a bounded height. The block fills it and scrolls internally.

Going live

The run function is a script that mutates step state in order. Replace it with a consumer of your agent’s event stream: map plan events to steps, tool events to the step’s tools, and emit a waiting step when the agent requests permission.