mcp-deepwiki

A connector skill for reaching DeepWiki, an MCP server that exposes documentation and AI-powered search over public GitHub repositories through its Ask Devin capability. It helps users understand open-source source code by reading generated repository documentation or asking questions directly about a project's code, without cloning or manually browsing the repository. The server is reached at its MCP endpoint, and calls are shown using npx and the mcporter CLI.

Three tools are documented. ask_question, the recommended path, takes a repository (or a list of up to ten, each in owner/repo format such as facebook/react) and a natural-language question, returning an AI-generated, context-grounded answer; because the answer is model-generated, the skill notes it can take a minute or more and advises setting a generous timeout. read_wiki_structure returns the list of documentation topics available for a repository, and read_wiki_contents returns the documentation itself. Each is accompanied by its JSON schema, making the expected arguments explicit - repoName for all three, plus question for ask_question.

The intended use is exploratory: point the tools at a public GitHub project to get an overview of its documentation structure, pull the underlying content, or pose targeted questions about how the code works and have them answered against the repository's context. Because it is a thin wrapper over the DeepWiki MCP endpoint, its capabilities are defined by that server rather than by local logic, and the owner/repo naming convention plus the multi-repository option on ask_question are the main things to keep in mind when invoking it. It is a good fit whenever the goal is to understand or interrogate an unfamiliar open-source codebase quickly.