context-budget

To audit Claude Code context window token usage, scan all loaded agents, skills, MCP servers, and project rules to estimate total context overhead. This identifies bloat and redundant elements, generating prioritized token-saving recommendations to optimize available context headroom and improve session stability.

Sluggish Claude Code session performance after adding MCP servers often stems from over-subscribed servers wrapping simple CLI tools, which consumes excessive context window tokens. Running a context audit detects this bloat and flags heavy components causing truncated AI outputs and reduced response quality.

The best way to check context capacity before expansion is to perform a pre-expansion capacity audit. This estimates current token overhead across agents, rules, and MCP servers, ensuring you have sufficient context headroom to add new components without causing truncated AI outputs.

Detecting redundant Claude Code components and verbose project rules requires a full component token inventory. This scan flags heavy agents, duplicate components, and verbose frontmatter, helping you pinpoint and remove context bloat to reclaim space for improved AI output quality.

You can optimize MCP servers and project rules to prevent truncated AI outputs by identifying over-subscribed servers and verbose configurations. A context audit generates prioritized recommendations to remove or lazy-load heavy components, reclaiming context space and stabilizing session performance.

You should run a context audit to catch Claude Code context creep after making changes to your environment. A post-change audit identifies new token overhead from updated agents, skills, or MCP servers early, allowing you to eliminate bloat before it impacts session performance and AI output quality.