MCP Builder
AI & MLby anthropics
2/3 audits passApache-2.0
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools.
MCP Builder, from Anthropic's skills repository, is a guide to building high-quality Model Context Protocol servers that let an LLM work with an external service. It runs in four phases. Phase 1 is research and planning: balance full API coverage against workflow tools, use clear prefixed tool names, keep responses concise with filtering and pagination, write actionable error messages, and study the MCP specification and SDK docs. It recommends TypeScript and streamable HTTP with stateless JSON for remote servers, stdio for local ones. Phase 2 is implementation of project structure, core infrastructure and tools; Phase 3 is review and testing; Phase 4 creates ten evaluation questions that test whether an LLM can actually use the server. References cover best practices and TypeScript and Python patterns.
Use it when wrapping an API or internal system as an MCP server in TypeScript or Python. It is about building servers, not about configuring an existing MCP client.
This is a coding-agent skill for tools like Claude Code; its evaluation scripts run locally. It is published under Apache-2.0 and importable into BusinessMCP, which already hosts a managed MCP endpoint for a workspace's data, so this skill matters most if you are writing your own server.
What you can do with it
- Wrap a REST API as an MCP server with clear tool names
- Choose streamable HTTP or stdio transport
- Write actionable error messages for agent-facing tools
- Create evaluation questions to test an MCP server
Run it on your business data
Imported into BusinessMCP, MCP Builder becomes a playbook your AI business analyst applies to your connected tools and MCP servers.
Use MCP Builder in BusinessMCPInstall it in a coding agent
One command adds MCP Builder to your project.
npx skills add https://github.com/anthropics/skills --skill mcp-builderHow we vetted it
- Source
- anthropics/skills at 3337550
- Licence
- Apache-2.0
- Security audits (skills.sh)
- Gen Agent Trust Hub: Pass · Socket: Pass · Snyk: Warn
- Bundled scripts
- Yes — read them before installing
Checked 2026-09-25 against its skills.sh listing. How we vet skills
Related skills
All skillsClaude API
Reference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file;
Skill Creator
Create new skills, modify and improve existing skills, and measure skill performance.
Find Skills
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities.
Frequently asked questions
Which language and transport does MCP Builder recommend?
TypeScript, with streamable HTTP and stateless JSON for remote servers and stdio for local servers. Python with FastMCP is also covered.
How does it test a finished server?
Phase 4 has the agent create ten evaluation questions that check whether an LLM can use the server's tools to answer realistic tasks.
How do I install MCP Builder?
Run `npx skills add anthropics/skills --skill mcp-builder`, or import it from the BusinessMCP dashboard.