context-budget

Claude Code context bloat typically stems from heavy MCP server schemas, redundant agent rules, or bloated CLAUDE.md files. Auditing your workspace components estimates token overhead and identifies which elements to remove or lazy-load to restore performance.

To audit token overhead, scan component directories to estimate per-file context costs using file heuristics and per-tool schema costs. Classify elements as keep, sometimes, or rarely-needed to generate a prioritized report targeting the highest-impact token savings.

Context budgeting is estimating and optimizing token consumption across agents, skills, rules, and MCP servers. You need it when debugging sluggish sessions after adding components or when planning capacity before expanding your workspace with more tools.

Yes, you can detect redundancy by scanning component directories and classifying elements as keep, sometimes, or rarely-needed. This flags overlapping components and bloated descriptions, producing a ranked report of immediate token-savings opportunities.

The best way to plan capacity is auditing existing context window consumption to estimate token overhead from current agents and servers. This generates a prioritized token-savings report, revealing whether you have budget to expand or need to optimize first.