{"version":"2.1.304","anchor":"context-usage-schema-adds-per-server-mcp-instructions-and-ag","canonical_anchor":"context-usage-schema-adds-per-server-mcp-instructions-and-ag","heading":"SDK context usage breakdown adds category ids, MCP instructions and paths","tier":"notice","area":"SDK","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.304\/e\/context-usage-schema-adds-per-server-mcp-instructions-and-ag","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.304","markdown":"### SDK context usage breakdown adds category ids, MCP instructions and paths\n\nThe SDK context usage report now gives categories stable ids and can list MCP instructions per server plus agent and skill paths\n\n**Unclear.** It is not clear whether these fields are new in this release or were already in the report before.\n\n**What**\n\nThe context window is the amount of text Claude can hold at once. Programs using Claude Code through the SDK, its programming interface, can ask for a breakdown of what fills it. That breakdown now carries more detail:\n\n- Each category has a stable id, such as `system_prompt`, `mcp_tools`, `free_space` and `autocompact_buffer`. Before, categories had only a name, a token count and a colour.\n\n- A new `localPaths` option adds `mcpInstructions`, the instructions each MCP server (an outside tool provider) adds, listed by server.\n\n- Agents and skills now include their `pluginId` and path.\n\n- These extra details appear only where Claude Code answers on its own standard input and output.\n\n**Why**\n\nTools built on the SDK can see in more detail what is using the context window and match categories reliably by id.\n\n- Area: SDK\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5\n- Scope: individual\n- Heads-up: no"}