How do I install the MCP Audit skill?

Run npx skillmds@latest add lucassantana-dev/mcp-audit in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is MCP Audit safe to use?

SkillMD's automated safety review verdict for this skill is PASS. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with MCP Audit?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is MCP Audit free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published MCP Audit?

LucasSantana-Dev (@lucassantana-dev) published this skill. Their other Agent Skills are listed on their SkillMD profile.

MCP Audit by lucassantana-dev — Agent Skill

Name: MCP Audit
Rating: 4.2 (5 reviews)
Author: LucasSantana-Dev

MCP Audit

Read-only diagnostic that scans Claude Code session transcripts to surface which MCP servers and tools you actually use, ranked by call frequency, with zero-use servers flagged for removal. Use when planning an MCP cleanup, evaluating whether to keep a newly-added server, deciding which servers warrant token cost in the catalog, or before authoring an MCP-removal PR. Outputs a markdown report (last N days) — does not modify any settings. Pair with the manual `claude mcp remove <name>` step once findings are reviewed.

LucasSantana-Dev Updated Aug 19, 2026 1 repo stars

File contents

mcp-audit

Read-only inventory of MCP-server tool usage from Claude Code session transcripts.

When to run

Before removing an MCP server (confirm zero use)
Periodic cleanup (suggest monthly via plugin-audit cron)
After adding a server, ~2 weeks in (validate it's earning its catalog spot)

How

bash ~/.claude/skills/mcp-audit/audit.sh [days]

days defaults to 30. Scans ~/.claude/projects/*/*.jsonl for mcp__<server>__<tool> invocations within the window, prints:

By server: count of tool calls per server, sorted desc
By tool: top 20 individual tools by call count
Zero-use servers: connected MCP servers (from claude mcp list) with no calls in window
Single-use servers: warning band (used 1–3 times — possibly tested-and-forgotten)

Output

Plain markdown to stdout. No mutations. No settings touched. Pipe to a file if you want history:

bash ~/.claude/skills/mcp-audit/audit.sh > ~/.claude/plans/mcp-audit-$(date +%Y-%m-%d).md

Acting on results

Zero-use server, ≥30 days connected → strong remove candidate. Confirm with claude mcp remove <name> -s user.
Single-use server, recent → leave in place (maybe newly added).
High-use server with low-quality tools → consider whether to disable specific tools rather than the whole server.

Limits

Counts tool invocations, not tokens or wall time. A single gh issue create call may be more valuable than 100 playwright_snapshot calls.
Doesn't see deferred tools that were never ToolSearch-loaded. If a server's tools are lazy and never reached, it shows as zero-use even if relevant.
Doesn't infer intent — a server may be unused because the workflow it supports hasn't come up recently, not because it's redundant.

---name: mcp-auditdescription: 'Read-only diagnostic that scans Claude Code session transcripts to surface which MCP servers and tools you actually use, ranked by call frequency, with zero-use servers flagged for removal. Use when planning an MCP cleanup, evaluating whether to keep a newly-added server, deciding which servers warrant token cost in the catalog, or before authoring an MCP-removal PR. Outputs a markdown report (last N days) — does not modify any settings. Pair with the manual `claude mcp remove <name>` step once findings are reviewed.'triggers: - mcp audit - which mcp servers do I use - audit mcp usage - find unused mcp - mcp cleanup - mcp removal candidates---# mcp-auditRead-only inventory of MCP-server tool usage from Claude Code session transcripts.## When to run- Before removing an MCP server (confirm zero use)- Periodic cleanup (suggest monthly via `plugin-audit` cron)- After adding a server, ~2 weeks in (validate it's earning its catalog spot)## How```bash ~/.claude/skills/mcp-audit/audit.sh [days]````days` defaults to 30. Scans `~/.claude/projects/*/*.jsonl` for `mcp__<server>__<tool>` invocations within the window, prints:1. **By server**: count of tool calls per server, sorted desc2. **By tool**: top 20 individual tools by call count3. **Zero-use servers**: connected MCP servers (from `claude mcp list`) with no calls in window4. **Single-use servers**: warning band (used 1–3 times — possibly tested-and-forgotten)## OutputPlain markdown to stdout. No mutations. No settings touched. Pipe to a file if you want history:```bash ~/.claude/skills/mcp-audit/audit.sh > ~/.claude/plans/mcp-audit-$(date +%Y-%m-%d).md```## Acting on results- Zero-use server, ≥30 days connected → strong remove candidate. Confirm with `claude mcp remove <name> -s user`.- Single-use server, recent → leave in place (maybe newly added).- High-use server with low-quality tools → consider whether to disable specific tools rather than the whole server.## Limits- Counts tool *invocations*, not tokens or wall time. A single `gh issue create` call may be more valuable than 100 `playwright_snapshot` calls.- Doesn't see deferred tools that were never `ToolSearch`-loaded. If a server's tools are lazy and never reached, it shows as zero-use even if relevant.- Doesn't infer intent — a server may be unused because the workflow it supports hasn't come up recently, not because it's redundant.

Frequently asked questions