Unclog: Identifies and fixes context bloat in Claude Code installations.

The modern AI developer stack, particularly those centered on complex code environments like Claude Code, runs into a fundamental limiting resource: the context window. While powerful, the sheer volume of auto-injected data—including historical context files, unused skills, cross-project duplicates, and configured but dormant Machine Context Providers (MCPs)—can rapidly bloat this window, leading to token wastage, performance degradation, and unexpected costs. Enter Unclog, a focused utility dedicated to diagnosing and remediating this 'context debt.' Unclog moves beyond simple cleanup scripts. Its core strength lies in its holistic auditing capacity. It systematically scans the entire Claude Code installation environment, providing granular reports on various sources of token consumption. This includes not just obvious file duplicates, but also subtle, often-missed context sources such as per-project `MEMORY.md` files, unused `Hooks`, and defunct remote MCP definitions. By categorizing these findings and calculating the exact token cost, the tool provides developers with a precise, actionable blueprint of their context bloat. Crucially, Unclog’s approach is built on extreme technical caution. Recognizing that developer workflows must remain uninterrupted, the tool does not operate in an 'auto-fix' mode. Instead, it generates a detailed report followed by a highly selective, user-gated repair interface. Developers can review each potential fix—be it curating an agent, disabling a skill, or removing a duplicate file—and apply changes piece by piece. Before any structural modification, a full snapshot of the current context is written, giving the user an immediate `unclog restore ` safety net. This level of explicit reversibility sets it apart from generalized cleanup utilities. For any advanced developer or DevOps team managing complex, multi-project AI pipelines, Unclog is less of an add-on and more of a vital maintenance utility. It is the specialized garbage collector for your AI context. By making the previously opaque problem of context window bloat visible, and providing a safe, reversible path to resolution, Unclog significantly improves the efficiency, predictability, and long-term maintainability of complex Claude Code deployments.