thomaschill

thomaschill / unclog

Public

Audit MCP servers, skills, and CLAUDE.md bloat eating your Claude Code context window.

pypi.orgprojectunclog claude claude-code claude-code-plugin cli context-window
19
1
100% credibility
Found Apr 23, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

unclog is a local command-line tool that scans Claude Code installations, displays token costs of installed agents, skills, commands, and MCP servers, and provides an interactive interface to selectively remove them.

How It Works

1
💭 Notice slow AI chats

Your conversations with your AI helper feel sluggish because it's carrying too much extra stuff in its memory from all the add-ons you've collected.

2
🔍 Find unclog

You learn about unclog, a friendly tool that helps clean up your AI helper's memory without any fuss or outside connections.

3
📥 Grab unclog

You add unclog to your setup with one quick step, like picking up a simple cleaning gadget.

4
🚀 Start unclog

You open unclog and it quietly checks everything in your AI helper's folder to see what's taking up space.

5
📊 Spot the space hogs

A colorful screen pops up showing your total memory use and lists each extra helper, shortcut, or tool connection with its exact cost, so you know exactly what's bloating things.

6
✂️ Pick and clear

You use simple keys to select the extras you don't need anymore, hit enter to confirm, and they vanish from your setup.

✨ Enjoy speedy chats

Your AI helper now starts faster with tons more room for your questions, saving you time and making conversations snappier every day.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is unclog?

Unclog is a Python CLI tool that audits your Claude Code installation for context-window bloat caused by agents, skills, slash commands, and MCP servers. Run `uv tool install unclog` followed by `unclog` to scan `~/.claude/`, see token costs via tiktoken, and use an interactive Rich picker to delete unused items—saving baseline tokens instantly. It's local-only with no telemetry, targeting macOS/Linux users on Claude Code 2.1.90+.

Why is it gaining traction?

In a sea of remote AI tools, unclog stands out as an mcp audit tool that unclogs your context without network calls, showing exact token savings from MCP audit logs and skills before/after cleanup. Devs love the one-command flow: scan, pick via keyboard (space toggle, a/A select sections), confirm destructive deletes. No alternatives offer this precise github audit trail for Claude's MCP servers and agent bloat.

Who should use this?

Claude Code power users bloated with 100+ agents/skills from experiments, or teams auditing MCP servers across projects to cut token costs. Ideal for AI-assisted coders hitting 200k limits on every prompt, needing an mcp audit github workflow without security audit mcp overhead.

Verdict

Grab it if you're deep in Claude Code—19 stars and 1.0% credibility signal alpha maturity with thin docs, but MIT-licensed code runs clean on Python 3.11+ and delivers real token wins. Test locally before production reliance.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.