Language:JapaneseEnglish

mcp-builder Claude Code Skill

Author
Anthropic
Topics
AI Agents & AI App Development
License
See LICENSE.txt
First cataloged
2026-07-09 (UTC)
Explanation last updated
2026-09-03 (UTC)
Source (GitHub) last updated
2026-04-20 (UTC) (138 days ago)

The explanation below is AI-generated. Please verify it against the sources.

It walks through a four-phase process: researching MCP design and protocol documentation, implementing the server (in TypeScript or Python), reviewing and testing the code, and finally creating evaluation questions to verify the server works well in practice. It gives concrete recommendations, such as favoring TypeScript with the official SDK and using streamable HTTP or stdio transports depending on deployment. It also links to supplementary reference files covering best practices, language-specific implementation patterns, and evaluation methodology. The guide is aimed at developers who want to create MCP tools with well-designed schemas, clear naming, and actionable error handling.

What you can do with mcp-builder

  • Follow a four-phase workflow (research, implementation, review/testing, evaluation) for building MCP servers
  • Get guidance on choosing between TypeScript (with Zod schemas) or Python (with Pydantic models) for implementation
  • Learn conventions for tool naming, input/output schemas, pagination, and error messages
  • Access reference guides for MCP best practices, language-specific patterns, and writing evaluation test questions
  • Learn how to test a built server using the MCP Inspector tool and verify build/compile correctness

Sources

Back to list