MCP Server Spec Writer skill for Claude and Codex
Specifies a Model Context Protocol server end to end, with tool schemas, auth and error cases, before any code is written.
Install it once and Claude or Codex runs it for you. When your request matches, your AI loads the skill, asks for anything it is missing, and works through the template itself. Prefer to fill it in by hand? See the full prompt template.
Download the skill (.zip) · View SKILL.md
When it is used
Use when an engineer is planning a Model Context Protocol server and needs tool definitions, schemas, auth and risks specified before coding. Do not use for implementing the server or choosing between agent frameworks. Related topics: AI, Engineering, Technical.
What it will ask you for
- System or data the server exposes (free text, e.g. "an internal CRM or a docs wiki")
- What an AI assistant should be able to do with it (longer free text)
- Access level (choose one: Read-only, Read and write, Write with confirmation)
- Candidate tools (longer free text)
- Authentication (choose one: API key, OAuth, None - local only)
- Include resources as well as tools (choose one: Yes, No)
- Rate limits, privacy rules or fields to redact (longer free text)
- Implementation language (choose one: TypeScript, Python)
How it works
- Map the example requests to a minimal set of tools, merging any that overlap.
- Write each tool's name and model-facing description stating when to use it and when not.
- Define input schemas with types and required fields, the return shape, and the error cases for each tool.
- Specify the auth flow, rate limits and redaction rules from the constraints.
- Write the risks section naming what write access could do if misused and the control that limits it.
What you get
A spec with an overview, one block per tool (name, description, input schema, return shape, errors), the auth flow, optional resources, and a short risks section.
What it watches out for
- Creating many narrow tools that confuse tool selection instead of a few clear ones.
- Writing tool descriptions for humans rather than for the model that must choose the tool.
- Returning entire records when a small summary would do, which wastes context.
How to install
Claude.ai and the Claude desktop app
- Download the skill .zip.
- Open Settings, then Capabilities, and find Skills.
- Choose Upload skill and select the .zip.
- Ask Claude for help with the task. It loads the skill when your request matches.
Claude Code
- Download the skill .zip.
- Unzip it into ~/.claude/skills/ (or .claude/skills/ inside a project).
- Start a new session. Claude Code loads the skill when your request matches.
Codex
- Download the skill .zip.
- Unzip it into ~/.agents/skills/ (or .agents/skills/ inside a project).
- Codex picks it up on its own. Type $ and the skill name to use it, or just ask for the task.
Cursor
- Download the Cursor rule (.mdc).
- Save it in .cursor/rules/ inside your project.
- Cursor's agent applies the rule when your request matches its description.
AGENTS.md (Codex, Cursor, and other coding agents)
- Copy the AGENTS.md section.
- Paste it into AGENTS.md at the root of your repository, creating the file if needed.
- Coding agents that read AGENTS.md follow it when your request fits.