context-budget

To audit token usage in Claude Code, run a context budget analysis to inventory agents, skills, rules, and MCP servers. This identifies heavy elements and produces a prioritized report with actionable recommendations to reclaim context window space.

Session performance degrades when new skills increase token overhead and cause context bloat. Auditing context usage detects overlaps and heavy elements across components, generating recommendations to rebalance the load and restore session responsiveness.

The best way to analyze MCP server token consumption is to run a context budget audit that evaluates .mcp.json configurations alongside agents and rules. This estimates token usage and flags heavy servers to ensure you maintain enough context headroom.

You can identify high token consumers by generating a Context Budget Report that inventories agents, skills, rules, and CLAUDE.md files. The report estimates token consumption for each element and flags the heaviest ones for prioritized optimization.

Yes, you can optimize CLAUDE.md to free up context window space by running a context budget audit. The analysis includes CLAUDE.md in its inventory to detect wasteful usage and overlaps, providing prioritized savings recommendations to reduce bloat.

You should run a context budget analysis when session performance falters, after adding new skills, or before scaling MCP servers. This ensures you identify wasteful token usage early and maintain adequate context headroom for future operations.