> ## Documentation Index
> Fetch the complete documentation index at: https://kit.near.tools/llms.txt
> Use this file to discover all available pages before exploring further.

# For AI Agents

> How to use near-kit documentation with AI coding assistants

This documentation is optimized for AI coding assistants. Whether you're using **Cursor**, **Windsurf**, **Claude**, **ChatGPT**, or other AI tools, there are multiple ways to give them context about `near-kit`.

<Warning>
  **Important:** `near-kit` is NOT `near-api-js`. AI models often confuse the two. Always provide context to ensure the AI uses the correct, modern API.
</Warning>

## Quick Access Methods

<CardGroup cols={2}>
  <Card title="Contextual Menu" icon="mouse-pointer">
    Select any text on this site and use the popup menu to send it directly to ChatGPT, Claude, Cursor, or VS Code.
  </Card>

  <Card title="MCP Server" icon="plug">
    Connect your AI tool to our Model Context Protocol server for live documentation access.
  </Card>

  <Card title="llms.txt" icon="file-text">
    Fetch the full documentation as a single text file optimized for LLMs.
  </Card>

  <Card title="Markdown Export" icon="download">
    Click the copy button on any page to get markdown for your AI context.
  </Card>
</CardGroup>

***

## Model Context Protocol (MCP)

MCP allows AI tools to connect directly to our documentation server. This gives your AI assistant live access to search and read the docs.

### Supported Tools

* **Claude Desktop** - Native MCP support
* **Cursor** - Via MCP extension
* **VS Code** - Via Copilot MCP extensions

### Connection URL

```
https://kit.near.tools/mcp
```

<Tabs>
  <Tab title="Claude Desktop">
    Add to your `claude_desktop_config.json`:

    ```json theme={null}
    {
      "mcpServers": {
        "near-kit": {
          "command": "npx",
          "args": ["-y", "@anthropic-ai/mcp-remote", "https://kit.near.tools/mcp"]
        }
      }
    }
    ```
  </Tab>

  <Tab title="Cursor">
    Add to your Cursor MCP settings:

    ```json theme={null}
    {
      "mcpServers": {
        "near-kit": {
          "url": "https://kit.near.tools/mcp"
        }
      }
    }
    ```
  </Tab>
</Tabs>

***

## llms.txt Files

The full documentation is available as plain text files optimized for LLM context windows.

<AccordionGroup>
  <Accordion title="llms-full.txt (Recommended)" icon="file">
    Contains the entire documentation in a single text file. Use this for most coding tasks.

    **URL:** [https://kit.near.tools/llms-full.txt](https://kit.near.tools/llms-full.txt)

    ```bash theme={null}
    curl https://kit.near.tools/llms-full.txt
    ```
  </Accordion>

  <Accordion title="llms.txt (Index Only)" icon="list">
    A condensed table of contents with page summaries. Use this if you're hitting context window limits.

    **URL:** [https://kit.near.tools/llms.txt](https://kit.near.tools/llms.txt)
  </Accordion>
</AccordionGroup>

***

## Using with AI Assistants

<Tabs>
  <Tab title="Cursor / Windsurf">
    1. Open the **Chat** pane
    2. Type `@Docs` and add the URL: `https://kit.near.tools/llms-full.txt`
    3. Or use `@Web` to fetch any documentation page directly

    The editor will ingest the documentation as a reference file.
  </Tab>

  <Tab title="Claude Desktop">
    Connect via MCP for the best experience:

    1. Open Claude Desktop settings
    2. Add the MCP server: `https://kit.near.tools/mcp`
    3. Claude can now search and read the docs directly

    Or paste the llms-full.txt URL in your conversation.
  </Tab>

  <Tab title="ChatGPT">
    1. Paste the URL in your chat: `https://kit.near.tools/llms-full.txt`
    2. Ask ChatGPT to read and reference it
    3. Add the recommended prompt below
  </Tab>

  <Tab title="VS Code Copilot">
    Use the contextual menu on this site:

    1. Select code or text you want to reference
    2. Click the VS Code option in the popup menu
    3. The content opens in your editor for Copilot context
  </Tab>
</Tabs>

***

## Recommended Prompt

When starting a new AI session, use this prompt to ensure accurate responses:

```text theme={null}
I am building an application using the `near-kit` TypeScript library for NEAR blockchain.

IMPORTANT: `near-kit` is NOT `near-api-js`. They are completely different libraries.
- Do NOT use `near-api-js` patterns (like `utils.format`, `account.functionCall`, `BN`)
- Use human-readable strings for amounts: "10 NEAR", "30 Tgas"
- Use the fluent transaction builder: `near.transaction().transfer().send()`
- Reference the provided documentation as the source of truth
```

<Note>
  This prompt helps prevent the AI from hallucinating `near-api-js` patterns, which is a common issue since that library has more training data.
</Note>