What do I need to use MCP Server Spec?

The product; The existing API surface; Who the agent acts for; The riskiest actions.

Is MCP Server Spec free to use?

Yes — run it free in the browser playground (no API key needed for the sponsored trial, or bring any provider key), or install it into Claude Code, Cursor, ChatGPT and more via npx pm-claude-skills. MIT-licensed and open source.

MCP Server Spec

Q: How do I know the MCP Server Spec output is good?

The skill verifies itself against its own quality checks: Every tool maps to an agent task; no tool exists because "the endpoint was there"; Any two sibling tools are distinguishable from their descriptions alone; Default responses fit comfortably in a context window (≤~2k tokens); Every destructive or outward-facing action is gated or on the never-list.

🔵 Stable🕐 updated 2026-07-02 🔷 SkillSpec L3 pm-agentnative

Design an MCP server for a product — the tool surface, auth model, and safety boundaries that make it genuinely usable by AI agents. Use when asked to spec an MCP server, expose a product to agents, design tools for Claude or other MCP clients, or review why an existing MCP server performs badly. Produces a complete server spec: a small task-shaped toolset with agent-tested descriptions, auth and scoping decisions, error design, and an explicit not-exposed list.

▶ Run it free — no key needed 📝 Grade your existing draft View SKILL.md ↗

What to give it

▸The product — and what users hire it for (the top 5 jobs, not the feature list)

▸The existing API surface — (endpoints or capability list) if one exists

▸Who the agent acts for — the end user's own account? a service account? multi-tenant?

▸The riskiest actions — the product supports (deletes, sends, payments, permission changes)

✅ The bar it holds itself to

Every skill in this library self-verifies — these are this skill's own quality checks, straight from its definition.

✓Every tool maps to an agent task; no tool exists because "the endpoint was there"

✓Any two sibling tools are distinguishable from their descriptions alone

✓Default responses fit comfortably in a context window (≤~2k tokens)

✓Every destructive or outward-facing action is gated or on the never-list

✓Errors name the offending parameter and the fix

✓The spec includes the agent-eval test plan, not just the schema

⚠️ What it refuses to do

Do not mirror the REST API — 40 endpoint-tools is the #1 way MCP servers fail

Do not write descriptions for developers ("wraps the /v2/items endpoint") — write them for a model choosing a tool

Do not return full API payloads — context windows are the scarce resource

Do not expose destructive actions ungated because "the client will be careful"

Do not skip the never-exposed list — an MCP server without one hasn't been threat-modelled

Install

npx pm-claude-skills add --agent claude   # or codex · cursor · gemini · hermes
# or one-line MCP (every skill, any client):
claude mcp add pm-skills -- npx -y pm-claude-skills-mcp

Related skills

🔌 Embed this skill

Drop this on your blog, docs, or site — it renders a "Run this skill" card:

<div data-pm-skill="mcp-server-spec"></div>
<script src="https://mohitagw15856.github.io/pm-claude-skills/embed.js" async></script>

💬 Discussion

MCP Server Spec is one of 599 open-source professional AI agent skills — all SkillSpec L3. Try them all in the browser · ⭐ Star on GitHub · Browse the full catalog