# DeepWiki MCP for GitHub repository documentation | FlyMy.AI

> Use DeepWiki MCP with your FlyMy.AI agent to find a topic in a public GitHub repository wiki, read its documentation, and ask a focused codebase question.

Page: https://flymy.ai/blog/deepwiki-mcp/

[Blog](https://flymy.ai/blog/) / DeepWiki MCP

# DeepWiki MCP for GitHub repository documentation

When you open a large repository, the first useful question is often “Where should I start reading?” DeepWiki MCP gives your FlyMy.AI assistant a way to navigate the wiki for an indexed public GitHub repository. Start with its table of contents, read the documentation, then ask about one decision in the code.

By **the FlyMy.AI team** 18 November 2026

![DeepWiki MCP workflow for a public GitHub repository: list its topics, read its wiki, or ask a codebase question.](https://flymy.ai/blog/deepwiki-mcp/img/og.jpg)

## Choose the right DeepWiki action

Use the outline when you need chapter names, the contents when you want to read the wiki, and a question when you can name the behavior you are trying to understand. An outline is usually enough to plan a first reading session. A question is more useful once you know the term the project uses.

**`read_wiki_structure`**

Get the topic outline for one repository. A maintainer onboarding a contributor can turn it into a reading order.

**`read_wiki_contents`**

Retrieve the wiki contents for one repository. Your assistant can summarize the material around a subsystem or collect terms for you to look up in the code.

**`ask_wiki_question`**

Send DeepWiki a written question about one repository or a list of repositories. It generates an answer grounded in their indexed documentation.

## Connect your assistant to DeepWiki MCP

1. Connect your assistant to FlyMy.AI and sign in. The [DeepWiki integration page](https://flymy.ai/mcp/deepwiki-public/#connect) has setup instructions for Claude Code, Codex, Cursor and other MCP clients.
2. Choose the public DeepWiki connector, `deepwiki_public`. Your FlyMy.AI sign-in is sufficient for the public wiki actions.
3. Give your assistant a repository name in `owner/repo` form. For example, `CognitionAI/deepwiki` identifies a repository. A link to an issue, branch or file points to a different resource.

## Find a starting point in a repository wiki

Ask for the outline of `CognitionAI/deepwiki`. That gives your assistant a list of topics from the wiki to work with. If you are preparing an onboarding note, ask it to identify a first chapter and explain why that chapter belongs at the beginning.

Try it

```
Use DeepWiki to list the wiki topics for CognitionAI/deepwiki and suggest where a new contributor should start.
```

## Read the GitHub repository documentation

For a second example, open [React’s public wiki](https://deepwiki.com/facebook/react). It has a topic on React DOM client rendering. Ask your assistant to read the wiki and select the sections you need. The contents action takes a repository name, so a topic such as React DOM belongs in your instructions to the assistant.

Try it

```
Read the facebook/react wiki with DeepWiki. Make a two-section reading list for someone working on React DOM, and explain why that order makes sense.
```

## Ask a focused codebase question

“How does this project work?” can produce an answer broader than your next decision. Give DeepWiki a specific event and ask what happens next. For a caching library, that event could be a component requesting data that is already cached.

Try it

```
Ask DeepWiki how TanStack/query decides whether to reuse a cached result or fetch it again. Identify the relevant concepts and say what I should check in the source.
```

To investigate a boundary between libraries, give the question action both repository names in one request. Asking it to label each side makes it easier to see which project to read or change. The question action accepts up to ten indexed public repositories.

Try it

```
Ask DeepWiki about facebook/react and TanStack/query together. When a cached query result changes, what is each project responsible for? Label the two sides separately.
```

## Check the answer against the source

For a reading list, keep the wiki section headings in your notes so someone else can follow the same path. For an implementation decision, follow the code references in the wiki and check the version you plan to ship. A wiki description can explain an architecture while the exact implementation changes between releases.

If a repository has no wiki, check its spelling and indexing on [DeepWiki](https://deepwiki.com/). If a question times out, its outcome can be uncertain. Check for a result before sending the same question again.

## Frequently asked questions

**What is DeepWiki MCP?**

It connects an MCP client to the wikis that DeepWiki maintains for indexed public GitHub repositories. In FlyMy.AI, the public connector can list wiki topics, read wiki contents, and ask DeepWiki a question about the codebase.

**Which GitHub repositories can I use?**

Choose a public repository with a wiki on DeepWiki. Supply its owner and repository name, for example facebook/react. If the repository has no wiki, choose an indexed repository for this workflow.

**Does DeepWiki MCP need a GitHub token?**

For these public wikis, sign in to FlyMy.AI and choose the public DeepWiki connector. It requires neither a GitHub token nor a DeepWiki API key.

**How do I ask about more than one repository?**

Give the question action an ordered list of up to ten indexed public repository names and one question. Read wiki outlines and contents one repository at a time.

For action arguments, response shapes and examples, see the [FlyMy.AI DeepWiki reference](https://docs.flymy.ai/mcp/deepwiki-public). The [DeepWiki MCP server documentation](https://docs.devin.ai/work-with-devin/deepwiki-mcp) describes the upstream service. FlyMy.AI integration updates appear in the [platform log](https://flymy.ai/changelog/).

---

For agents: the FlyMyAI operating guide (sign up, connect, run models, tools and agents) is https://flymy.ai/skill.md, and https://flymy.ai/llms.txt lists the other pages and the documentation.
