yWorks/mcp-typescribe
Free高效地向代理提供 TypeScript API 信息，使其能够处理未经训练的 API。
FreeFree tier
About yWorks/mcp-typescribe
An open-source MCP (Model Context Protocol) server implementation that enables LLMs to work with new TypeScript APIs and frameworks not seen during training. It loads TypeDoc-generated JSON documentation and indexes it for efficient querying, exposing a set of query endpoints that allow AI agents to search for symbols, get detailed information about specific API parts, and understand relationships between components. The project aims to solve the bottleneck of LLMs relying solely on training data for API knowledge, enabling plug-and-play support for unfamiliar APIs without retraining.
Key Features
Parses TypeScript definitions into machine-readable format via TypeDoc integration
Serves context dynamically to LLMs through MCP protocol (Claude, Cline, Cursor, Windsurf)
Comprehensive query capabilities: search for symbols, get detailed API info, understand relationships
Loads and indexes TypeDoc-generated JSON documentation for efficient querying
Enables agentic behavior: LLMs can query, plan, and adapt to unfamiliar APIs without retraining
Pros & Cons
Pros
- Allows LLMs to work with APIs they were never trained on, overcoming a key limitation
- Efficiently queries large API documents without putting everything in context
- Open source and free to use with any MCP-compatible tool
- Easy integration with popular AI coding assistants
- Designed for scalability: loads and indexes documentation once, then queries dynamically
Cons
- Currently only supports TypeScript type definitions (other languages not yet implemented)
- Public development has been suspended as no active community formed; focus shifted to yFiles-specific MCP server
- Requires users to generate TypeDoc JSON documentation from their TypeScript code first
Best For
Enabling LLM-based coding assistants to use new or internal TypeScript APIsFaster onboarding for proprietary SDKs or libraries not in training dataPowering more autonomous, context-aware coding agents that explore unknown APIsProviding real-time API documentation context to AI tools like Claude, Cline, Cursor, or Windsurf
FAQ
What is MCP-Typescribe?
It's an open-source MCP server that provides LLMs with real-time access to TypeScript API documentation, allowing them to work with unfamiliar APIs by querying TypeDoc-generated data.
Why was development suspended?
The project is on hold due to lack of active community engagement. The authors are working on a follow-up project specifically for the yFiles API.
What languages or file formats does it support?
Currently it only works with TypeScript type definitions via TypeDoc JSON output. Other languages are not yet supported.