{
  "id": 14253091,
  "title": "Claude usage jumped? How I added project, session and subagent attribution to my usage tracker",
  "url": "https://urgent.news/2026/10/13/claude-usage-jumped-how-i-added-project-session-and-subagent",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-13T17:28:52.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/rehberodhano/claude-usage-jumped-how-i-added-project-session-and-subagent-attribution-to-my-usage-tracker-5bjm"
  },
  "original_language": "en",
  "account": "Claude's usage tracker, initially released as the Claude Usage Companion, has recently been upgraded with new features to provide more detailed insights into user activity. The tracker now includes session and subagent attribution, allowing users to identify why their weekly usage bars have jumped unexpectedly. This feature aims to help users understand their usage patterns better, whether it's due to a month-long Claude Code session or a subagent running longer than anticipated.\n\nThe updated tracker consists of three main tabs: Overview, Activity, and Sessions. The Overview tab provides a high-level view of the user's progress, including the reset times and history. It also features a forecast indicating when the limit will be reached, along with a projection of how much earlier than the reset this will occur.\n\nThe Activity tab breaks down Claude Code usage by project, model, and session. It offers a more granular view of the user's usage, showing which projects are consuming the most resources and how the model mix is distributed. Additionally, the daemon connection allows for the tracking of subagents within each session, displaying the type, description, model, and token usage for each subagent separately.\n\nThe Sessions tab provides a comprehensive list of all previous Claude Code sessions, which can be searched and resumed using the daemon. This feature ensures that users never lose valuable work that may have been running in the background. Unlike the other tabs, the Sessions tab does not automatically appear until a daemon connection is established.\n\nClaude's usage tracker uses a sustainable pacing calculation to estimate the average usage rate needed to avoid hitting the reset limit. This calculation takes into account the amount of usage remaining and the hours left until the reset. The tracker also includes an anomaly detection feature that flags any sudden spikes in usage, comparing the current rate to the user's recent activity patterns. When an anomaly is detected, a separate Usage Inspector provides more details about the session during that spike, including the largest local contributor, model mix, and session usage.\n\nIn summary, the Claude Usage Companion has evolved to give users a more detailed understanding of their usage patterns, allowing them to optimize their usage, identify anomalies, and avoid unexpected resets.",
  "summary": "The first version of Claude Usage Companion covered the limit bars, a burn-rate forecast, and basic project and model attribution from Claude Code. That told me where I stood. It didn't tell me why. What kept happening was this: I'd open Settings → Usage and the weekly bar had jumped. Was it yesterday's refactor? One long Claude Code session? Did a model dominate the week? Did a subagent run much…",
  "key_points": [
    "Claude Usage Companion upgraded with session and subagent attribution",
    "Three main tabs: Overview, Activity, and Sessions",
    "Anomaly detection flags sudden spikes in usage"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "XDA Developers",
        "title": "5 ways you're wasting Claude usage limits without even realizing it",
        "url": "https://urgent.news/2026/10/13/5-ways-youre-wasting-claude-usage-limits-without-even-realizing-it",
        "published": "2026-10-13T20:00:17.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}