{"version":"2.1.287","anchor":"token-usage-breakdown-prompt-builder-with-no-visible-caller","canonical_anchor":"token-usage-breakdown-prompt-builder-with-no-visible-caller","heading":"The built-in request to break down a session's token use has been reworked","tier":"notice","area":"Context","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287\/e\/token-usage-breakdown-prompt-builder-with-no-visible-caller","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287","markdown":"### The built-in request to break down a session's token use has been reworked\n\nA built-in request asking Claude where a session's tokens went now groups usage by source and asks for a chart with caveats\n\n**Unclear.** It is not clear how this request is reached or which of its parts differ from the earlier version.\n\n**What**\n\nClaude Code contains a ready-made request, \"Show me where this session's tokens went.\", that asks the model to explain where a session's tokens were spent. Tokens are the units of text the model reads and writes, and they are what usage is measured in. This request has been changed. It now:\n\n- groups the measured usage by tool, MCP server, instructions, person, Claude and summary\n\n- counts text read back from the cache at about a tenth of its size and the model's own output at about five times its size\n\n- asks the model for a chart along with notes on what the numbers do not show\n\n**Why**\n\nNo command or menu that sends this request has been identified, so there is no known way to use it yet.\n\n- Area: Context\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5"}