Cursor has three built-in documentation tools to inject precise context into AI programming

In previous articles, we have introduced that due to the inherent reasons of large models, it is impossible to master the latest frameworks and libraries, so old and outdated interfaces often appear in programming, or a non-existent interface is created, resulting in incorrect code generation. Therefore, effectively utilizing various document resources to provide high-quality, real-time context for AI is the key to overcoming this challenge and fully realizing its potential.

For this purpose, we introduced tools such as context7.

MCP Server! " data-itemshowtype="0" linktype="text" data-linktype="2">Must install! The magical MCP Server that effectively reduces Cursor illusion and improves the quality of code generation!

In fact, similar features have long appeared in Cursor, but because they are not easy to use and are not known how to use, they have not been taken seriously by developers. Today, I will bring you the official suggestions of Cursor in this regard ^[1] to help you better utilize the document interaction features to provide accurate context for AI, thereby improving the accuracy of large model programming.

Cursor provides three tools for processing external public documents and private internal documents respectively.

1. Use @Docs：Get authoritative official information

@Docs The Cursor function connects you directly to the official documentation for many popular tools and frameworks. When you need authoritative information of the following types,@Docs is preferred:

• API Reference : Find precise function signatures, parameter definitions, return value types, and more.
• Getting Started Guide : Quickly learn how to install, configure, and use the tool.
• Official Best Practices : Get recommended coding patterns and design principles from the source.
• Framework-specific debugging : Refer to the official troubleshooting steps and FAQs.

pass @Docs, developers can obtain and apply the most reliable technical information directly in the Cursor environment.

Note: Before use, you need to add commonly used documents to the cursor.

2. Use @Web: Explore a wide range of community and online resources

@Web The Cursor function enables real-time searching of the Internet to obtain the latest blog posts, community discussions, and tutorials. @Web:

• Recent Tutorials and Code Examples : Find the latest learning resources contributed by the community.
• Comparison of technical solutions : Read articles analyzing the pros and cons of different methods or tools.
• Stay up to date : Learn about very recent technology updates, version releases, or industry announcements.
• Get multiple perspectives : See multiple solutions and viewpoints on the same problem from different developers.

@Web It expands the boundaries of information acquisition, which is particularly suitable for rapidly changing technology fields and scenarios that require multiple perspectives.

3. Use MCP to access internal documents

Internal enterprise documents, such as internal API specifications, company coding standards, proprietary system descriptions, and specific business logic, are not accessible to AI models during standard training. This information is critical to ensuring project quality and team collaboration. The Model Context Protocol (MCP) is a set of standardized methods provided by Cursor that aims to safely and efficiently integrate an enterprise's private documents and internal systems into the cognitive scope of AI. MCP acts as a bridge between Cursor and internal enterprise resources.

For enterprises with special internal systems or unique document management needs, you can build a custom MCP server. Here is an example (context7 is actually a similar implementation):

import { McpServer, ResourceTemplate } from  "@modelcontextprotocol/sdk/server/mcp.js" ;
import { StdioServerTransport } from  "@modelcontextprotocol/sdk/server/stdio.js" ;
import { z } from  "zod" ;
import TurndownService from  "turndown" ;

// Create an MCP server  for  scraping internal docs
const server = new McpServer({
  name:  "internal-docs" ,
  version:  "1.0.0"
});

const turndownService = new TurndownService();

// Add tool to scrape internal documentation
server.tool( "get_doc" ,
  { url: z.string() },
  async ({ url }) => {
    try {
      const response = await fetch(url);
      const html = await response.text();
      
      // Convert HTML to markdown
      const markdown = turndownService.turndown(html);
      
      return  {
        content: [{  type :  "text" , text: markdown }]
      };
    } catch (error) {
      return  {
        content: [{  type :  "text" , text: `Error scraping  ${url} :  ${error.message} ` }]
      };
    }
  }
);

// Start receiving messages on stdin and sending messages on stdout
const transport = new StdioServerTransport();
await server.connect(transport);

This allows companies to flexibly access resources such as internal websites, proprietary databases, custom document systems or internal knowledge bases based on their own circumstances.

In addition, the value of documents lies in their accuracy and timeliness. Cursor not only supports the consumption of existing documents, but also has the ability to assist developers in creating and updating documents, promoting the accumulation and sharing of knowledge. The official also gave two suggestions for this:

1. Generate documentation from code : Developers can select code snippets and use Cursor to quickly generate comments, functional descriptions, or preliminary API documentation frameworks to reduce the burden of document writing.

Please generate a standard documentation string (docstring) for the selected Python function, describing its functionality, parameters, and return value.

1. Extract and structure information from the conversation : After collaborating with Cursor to solve a complex problem or complete a design, you can instruct Cursor to summarize the key information, decision logic, and code examples in the interaction process into a structured document or memo. For example, you can prompt:

Please organize our discussion about [specific module] into a technical description, including the main design points, problems encountered, solutions, and relevant code examples.

summary

Models are becoming more and more powerful and can correctly complete a programming task. In many cases, the model is not the bottleneck. The key lies in whether it can provide high-quality context. @Docs,@Web and MCP It provides developers with a powerful and flexible document interaction toolset, supplemented by dynamic document generation capabilities, which effectively bridges the gap between the inherent knowledge of AI models and the actual needs of projects. By proficiently using these advanced document techniques, developers can significantly improve the efficiency and quality of collaboration with Cursor, which is also one of the important criteria for measuring the level of programmers in the future.