Deepwiki.com:MCP - chunhualiao/public-docs GitHub Wiki

Yes, deepwiki.com does provide an MCP server for open source projects indexed on the site, including the ROSE compiler project at https://deepwiki.com/rose-compiler/rose-full. This is a free, remote service (no authentication required) that offers programmatic access to repository documentation and AI-powered tools for context gathering and question answering via the Model Context Protocol (MCP). It includes three main tools:

read_wiki_structure: Retrieves a list of documentation topics for a GitHub repository (e.g., rose-compiler).
read_wiki_contents: Views specific documentation content for a repository or page (e.g., the rose-full topic).
ask_question: Submits any question about a repository and receives an AI-generated, context-grounded response.

The base server URL is https://mcp.deepwiki.com/, with endpoints supporting SSE (Server-Sent Events) at /sse (recommended for most integrations) or Streamable HTTP at /mcp.

OpenAI example

https://platform.openai.com/docs/guides/tools-connectors-mcp

from openai import OpenAI

client = OpenAI()

resp = client.responses.create(
    model="gpt-5",
    tools=[
        {
            "type": "mcp",
            "server_label": "deepwiki",
            "server_url": "https://mcp.deepwiki.com/mcp",
            "require_approval": {
                "never": {
                    "tool_names": ["ask_question", "read_wiki_structure"]
                }
            }
        },
    ],
    input="What transport protocols does the 2025-03-26 version of the MCP spec (modelcontextprotocol/modelcontextprotocol) support?",
)

print(resp.output_text)

How to Test the MCP Server for https://deepwiki.com/rose-compiler/rose-full

Since this is a remote MCP server designed for integration with AI coding assistants (e.g., Cline in VS Code, Claude Desktop, Cursor, or Devin), testing typically involves configuring your client to connect to the server and then invoking its tools via prompts or API calls. No local installation is needed. Here's how to set it up and test:

Configure the MCP Server in Your Client:

Users > liao6 > Library › Application Support › Code > User > globalStorage › saoudrizwan.claude-dev › settings > f} cline_mcp_settings.json

For Cline or Similar VS Code Extensions (e.g., Cursor): Open your configuration file (often .cursor/mcp.json or equivalent in Cline's settings). Add the server like this:
```
{
  "mcpServers": {
    "deepwiki": {
      "serverUrl": "https://mcp.deepwiki.com/sse"
    }
  }
}
```

Invalid MCP settings schema by Cline -> MCP Servers-> Conifgure

 Restart VS Code or reload the extension. Access it via the Cline panel (sidebar icon > menu > MCP Servers) to connect and manage.

For Claude Desktop/Code: Use the command:

claude mcp add -s user -t http deepwiki https://mcp.deepwiki.com/mcp

This setup allows the AI agent to invoke the DeepWiki tools during tasks.

Test with Tool Invocations:
- In your AI agent's prompt interface (e.g., Cline's chat or agent mode), reference the server and test the tools. Examples:
  - To get the structure: Prompt something like "Using DeepWiki MCP, read_wiki_structure for rose-compiler repository."
  - To fetch content: "Using DeepWiki MCP, read_wiki_contents for https://deepwiki.com/rose-compiler/rose-full."
  - To ask a question: "Using DeepWiki MCP, ask_question: What is the ROSE-full architecture in the rose-compiler project?"
- The agent will send requests to the MCP server and incorporate responses into its workflow.
Direct API Testing (for Advanced Verification):
- Use curl or a similar tool to simulate an MCP request. For example, to test read_wiki_contents for the rose-full page:
```
curl -X POST https://mcp.deepwiki.com/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "action": "read_wiki_contents",
    "params": {
      "url": "https://deepwiki.com/rose-compiler/rose-full"
    }
  }'
```
  Expect a response with Markdown-formatted documentation content. Adjust the endpoint to /sse if using Server-Sent Events.

If you encounter issues (e.g., due to anti-scraping measures on unofficial servers), stick to the official endpoints above. For more details, refer to the official documentation.