MCP-Typescribe

MCP-Typescribe is an MCP server that provides AI coding assistants with targeted access to TypeScript API definitions using TypeDoc JSON exports. It connects Large Language Models to local API documentation structures, allowing software engineers and developers using MCP clients like Cline, Cursor, or Claude to query unknown or proprietary libraries. Instead of loading an entire codebase or large definition files directly into the LLM context window, MCP-Typescribe indexes structured documentation and offers specialized search tools. This enables agents to inspect class members, view inheritance hierarchies, look up symbol signatures, evaluate parameter details, and inspect JSDoc descriptions on demand. By querying only the relevant segments of an API during development tasks, the server allows AI models to write accurate code against internal SDKs and new APIs without requiring training data or exhausting token limits.

Category: Developer Tools & Code Intelligence

Tags: code analysis, documentation, typedoc, typescript

Visit MCP-Typescribe

How to install and configure MCP-Typescribe

1. Generate TypeDoc JSON documentation for your project by running npx typedoc --json docs/api.json --entryPointStrategy expand path/to/your/typescript/files (or configure via tsconfig.json). 2. Build the project locally if installing from source with npm install and npm run build. 3. Add the server configuration to your MCP client settings (for example, in cline_mcp_settings.json): json { "mcpServers": { "typescribe": { "command": "npx", "args": [ "-y", "mcp-typescribe@latest", "run-server", "<PATH_TO_API_DOT_JSON>" ], "env": {} } } } 4. Replace <PATH_TO_API_DOT_JSON> with the path to your generated TypeDoc JSON file and restart your MCP client.

What you can do with MCP-Typescribe

• Search TypeScript symbols by name and kind to identify relevant classes, interfaces, and methods in unfamiliar SDKs. - Query parameter details and return types to accurately construct function calls without consulting external documentation. - Inspect class hierarchies and inheritance trees to find all implementations of specific interfaces or extended classes. - Retrieve JSDoc comments and description metadata across an API to understand developer usage instructions and edge cases. - Discover where specific types or functions are referenced across a codebase to support large-scale refactoring tasks.

Key facts

• https://github.com/yWorks/mcp-typescribe
• Developer Tools & Code Intelligence, Notes & Knowledge Management
• code analysis, documentation, typedoc, typescript

Part of MCP Servers

How do I install MCP-Typescribe?

You can run MCP-Typescribe directly through npx using the mcp-typescribe@latest package. First generate a TypeDoc JSON file for your TypeScript project using typedoc, then register the server in your MCP client configuration by passing npx as the command, along with run-server and the path to your generated JSON documentation file as arguments.

What can MCP-Typescribe do?

MCP-Typescribe provides tools for AI assistants to explore TypeScript APIs. It allows agents to search symbols, inspect detailed symbol specifications, list class members, retrieve function parameter information, locate interface implementations, search by return type or JSDoc descriptions, check type hierarchies, and discover symbol usages without reading raw source files.

Which MCP clients work with MCP-Typescribe?

MCP-Typescribe adheres to the standard Model Context Protocol and works with compatible AI clients such as Cline in VSCode, Cursor, Windsurf, Claude Desktop, and the official MCP inspector tool.

Is MCP-Typescribe open source?

Yes, MCP-Typescribe is an open-source project published under the MIT license and created by yWorks GmbH. The source code is publicly hosted on GitHub.