MCP Server Audit

    1

    Audits connected MCP servers by tool-definition token cost and actual usage, so you can disconnect dead weight.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    MCP Server Audit

    Example session with this skill installed

    I have a lot of MCP servers connected. /mcp-audit — which ones are costing me tokens for nothing?

    • Read your context and instructions
    • Compiled the mcp server audit
    🔌 3 MCP server(s) loaded — ~13,400 tokens in tool definitions
    
    Server       Tools  Est. Tokens  Used This Session  Relevant?
    filesystem   8      ~2,400       ✅ 3 tools used    ✅ Yes
    github       15     ~6,200       ❌ None            ⚠️ Maybe
    slack        12     ~4,800       ❌ None            🔴 No
    
    🔴 Disconnect these — not used, not needed
    - slack (~4,800 tokens) — no messaging actions in this task.
      Run: /mcp disconnect slack
    
    🟡 Questionable — loaded but unused
    - github (~6,200 tokens) — you haven't opened any PRs yet, but this task may involve one later.
    
    🟢 Keep — actively used or clearly needed
    - filesystem
    
    💡 Optimization moves
    - Disconnecting slack recovers ~4,800 tokens immediately
    - Consider setting github to manual-connect if you only use it occasionally
    

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Inventories all connected MCP servers, estimates the token cost of their tool definitions, checks which tools have actually been called this session, and flags servers that are costing tokens without contributing to your current task.

    When to use

    • You have many MCP servers configured and want to know what they cost
    • Context is getting full and you want to reclaim token budget
    • You want to audit which servers you actually use in a typical session

    How it works

    Invoke it by name in your agent (in Claude Code: /mcp-audit). It reads only the MCP tool definitions and tool calls already in the current session — it doesn't call any MCP tools, write files, or make network calls. You decide what to disconnect.

    Part of Claude Token Inspector, an open-source (MIT) suite of 8 context-window skills by the original author. Source: https://github.com/msishariful/claude-token-inspector

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    Token counts are heuristics-based estimates. Does not support automated server disconnection; requires manual execution of commands.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 23 days ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed23 days ago

    What's inside

    Frequently Asked Questions

    This skill acts as a diagnostic layer for your MCP environment, scanning active tool definitions and JSON schemas to identify which servers are consuming the most space in your context window. It solves the problem of 'token bloat' by pinpointing exactly which tools are idling while draining your budget.

    It's a standard SKILL.md skill, built and tested on Claude Code. It works in other SKILL.md-compatible agents that load MCP servers (Cursor, Codex CLI, Gemini CLI, etc.), though the suggested /mcp commands are Claude Code conventions.

    No. It only reads what's in the session and writes a report. You decide what to disconnect and run the commands yourself.

    It's an estimate based on the size of each server's tool definitions (roughly 1 token per 3–4 characters of JSON schema), not your billing data. Use it to compare servers and spot the expensive ones.

    No setup beyond installing the skill; run it by name (in Claude Code: /mcp-audit). You get the SKILL.md instructions file. It's free and open source (MIT) and part of the Claude Token Inspector repo on GitHub.

    New in LLM Ops

    Free