Skip to Content
CardsAgent Pulse

Agent Pulse card

Agent Pulse shows how your agents spend their time. It is a verified custom card by NeuroSquad: in the add menu (+ on the canvas) pick Custom card…, install Agent Pulse from the Verified tab and add it to the canvas.

Requests

Draw arrows between the card and the agents you want to compare — for example Claude Code, OpenCode and Codex running the same prompt on the same model. Each agent gets a lane on one time axis, and every model request is a bar from the moment it was sent to the moment the response completed.

  • Bar length — how long the request took; bar height — its output tokens. Hover or focus a bar for the duration, model, tokens and the tool calls the model asked for.
  • A faint band shows when the agent was working (amber — waiting for you); ▼ marks a prompt, a green line the finish.
  • Per lane: number of requests, total model time, median and longest request. The KPIs on top are over all requests; zoomed out, the card’s tile shows requests and the median request time.
  • Run fits the whole run — agents run one after another sit side by side in time, simultaneous ones overlap. 15m / 1h / 4h follow the clock. Expanded, scroll to zoom and drag to pan; Fit resets.

The data comes from each agent’s own log, the same as the Usage section. Most harnesses record when a request was sent (Claude Code, Codex, OpenCode, Kilo Code, pi, omp, Kimi Code, Copilot CLI, Gemini CLI, Cline, Qwen Code, Crush); Hermes, Factory Droid, Goose, aider and Auggie record only when it completed, so their requests are ticks rather than bars. Amp and Cursor keep no log NeuroSquad can read. The view needs NeuroSquad 0.1.257 or newer and asks for the optional token usage permission the first time.

Status

The Status view is the card’s original picture of the whole canvas: a timeline per agent (working, waiting for you, finished, idle), turns and their length, and how long agents waited on you — with a gentle nudge when one waits too long.

Good for: benchmarks (with Run Stats next to each agent), and seeing which agent is slow, busy or blocked on you.