unclog
Identifies unused or bloated MCP server configurations and skill definitions that consume Claude's context window, freeing capacity for actual task completion.
Command-line tool that scans your Claude Code install and shows what every agent, skill, slash command, and MCP server costs in context-window tokens on every turn, including a baseline total loaded before you type anything. It flags MCP servers with zero invocations in the last 30 days and opens an interactive picker to delete what you don't use — locally, with no telemetry, accounts, or network calls.
- Identify unused MCP servers consuming context in production deployments
- Detect oversized CLAUDE.md files that could be pruned
- Audit skill definitions before scaling Claude integrations
Identifies unused or bloated MCP server configurations and skill definitions that consume Claude's context window, freeing capacity for actual task completion.
Teams running multiple MCP integrations who need to audit which tools are actually in use versus draining context unnecessarily.
https://github.com/thomaschill/unclog
By thomaschill
How to Get It
uv tool install unclog
Tip: Paste this into a Claude Code conversation. Verify command matches your Claude Code version.
After installing, paste this into Claude:
Help me identify unused MCP servers consuming context in production deployments
Trust Signals Auto-scanned
Community Pulse Emerging
Discussed on Hacker News
- Show HN: Unclog – find and fix Claude Code context bloat — Hacker News · 3 pts
1 mentions across 1 sources
Reviewer notes
Auto-scanned review. These are observations, not a security certification.
Scored from trust signals (evidence-eval-v1): 31 GitHub stars; contributors unknown; last commit 40d ago; license MIT.
Things to check
- Removal actions are destructive: files are deleted from disk and MCP servers removed from ~/.claude.json with no undo and no snapshot — restore from your own backup or reinstall. It cannot show per-MCP token costs (session logs no longer record tool schemas), so it shows 30-day invocation counts instead.
- Single maintainer. Consider the risk if this person stops maintaining the project.
How to evaluate tools before deploying →
Data shown here comes from public APIs and automated scanning. Reviewer notes reflect one person's experience. This is not a security certification or legal recommendation. Always evaluate tools according to your own organization's policies.
Evaluation
Scored from trust signals (evidence-eval-v1): 31 GitHub stars; contributors unknown; last commit 40d ago; license MIT.