Mcp Audit
Run free context-engineering audits on your agent setup with your own Anthropic key and get the top three concrete findings per scan.
Overview
io.github.archonics/mcp-audit is an MCP server for the Ship phase that runs context-engineering audits on AI agents and returns the top three findings per scan using your Anthropic API key.
What is this MCP server?
- Free context-engineering audits for AI agents (BYOK Anthropic API key)
- Top-3 findings per scan for fast prioritization
- stdio npm package @archonics/mcp-audit v0.1.4 runs audits from your machine
- Audit engine calls Claude directly with your ANTHROPIC_API_KEY—no opaque third-party model routing
- GitHub: archonics/mcp-audit
- Top-3 findings per scan (documented product limit)
- Server version 0.1.4; npm package @archonics/mcp-audit 0.1.4
- Single required env var: ANTHROPIC_API_KEY
What problem does it solve?
Agent stacks grow messy context—redundant tools, vague instructions, and risky patterns—and solo builders lack a quick audit pass before shipping.
Who is it for?
Builders shipping Claude-powered agents who want a lightweight, BYOK audit loop without a paid SaaS audit platform.
Skip if: Teams that need full SOC2-style assessments, binary SAST on repos, or audits without any Anthropic API spend.
What do I get? / Deliverables
Each scan surfaces three prioritized context-engineering findings so you can tighten prompts, tools, and workflows before users depend on them.
- Context-engineering audit report with top-3 findings per scan
- Actionable priorities for prompt and tool configuration
- Repeatable local audit workflow via MCP
Recommended MCP Servers
Journey fit
Context audits belong on the Ship shelf under review because you run them before you trust an agent workflow in production or after you change prompts and tools. review captures systematic checks on how context is assembled—not penetration testing of a web app, but engineering review of agent context design.
How it compares
Agent context audit MCP, not a repository code scanner or a generic LLM linter skill.
Common Questions / FAQ
Who is mcp-audit for?
Solo builders and small teams running Anthropic-based agents who want structured context-engineering feedback before or after shipping agent changes.
When should I use mcp-audit?
Run a scan after you add MCP servers, reshape system prompts, or before launch when you need the top three context risks spelled out.
How do I add mcp-audit to my agent?
Install @archonics/mcp-audit from npm, configure stdio MCP in your client, and set the required secret environment variable ANTHROPIC_API_KEY.