{"version":"2.1.282","anchor":"no-cache-breakpoint-on-api-system-messages-under-claude-code","canonical_anchor":"no-cache-breakpoint-on-api-system-messages-under-claude-code","heading":"Simulating proxy usage now also turns off caching markers on system messages","tier":"use","area":"Elsewhere","url":"https:\/\/changelogs.core-directive.com\/v\/2.1.282\/e\/no-cache-breakpoint-on-api-system-messages-under-claude-code","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.282","markdown":"### Simulating proxy usage now also turns off caching markers on system messages\n\nWith `CLAUDE_CODE_SIMULATE_PROXY_USAGE` set, Claude Code no longer puts caching markers on system messages\n\n**What**\n\nA cache breakpoint is a marker that tells the API which part of a request it can reuse from an earlier one. When the environment variable `CLAUDE_CODE_SIMULATE_PROXY_USAGE` is set, Claude Code now also leaves these markers off system messages sent to the API. Before, only an existing check controlled that.\n\nThe step that places these markers also now accepts a message from which the conversation branches off.\n\n**Why**\n\nThis mostly matters to people testing how Claude Code behaves behind a proxy.\n\n- Area: Elsewhere\n- Names: `CLAUDE_CODE_SIMULATE_PROXY_USAGE`\n- Tier: Use it now\n- Useful: 2\/5\n- Signal: 1\/5"}