Skip to main content

AI tools

Skills​

SVAR skills work with AI coding agents that support the SKILL.md convention, including Claude Code, Cursor, OpenAI Codex CLI, and Gemini CLI.

Run the following command to install the skill:

npx skills add svar-widgets/skills --skill svar-svelte

You can also download the skill folder manually from the skills repository.

Using skills​

Skills are loaded automatically when the agent decides they match the task. To guarantee that a skill is used, invoke it explicitly:

  • In Claude Code: start your prompt with /svar-svelte
  • In OpenAI Codex CLI: start your prompt with $svar-svelte

Flat context files for other tools​

If you use tools without native skill support, like ChatGPT, GitHub Copilot, or browser-based playgrounds, the guides folder provides the same context as plain markdown files you can paste or attach:

  • svar-svelte-<package>.md — context for a single package: svar-svelte-core.md, svar-svelte-comments.md, svar-svelte-menu.md, svar-svelte-tasklist.md, svar-svelte-toolbar.md

MCP​

In the modern development landscape, AI automation tools have become essential for efficient coding and problem-solving. To support this, we provide dedicated resources that allow AI agents to better understand and work with our products.

SVAR MCP​

Currently, we offer a Model Context Protocol (MCP) server, which connects your AI assistant directly to our latest documentation and code examples.

The SVAR MCP server acts as a bridge between our documentation and your AI coding tool. Large Language Models (LLMs) often have training data cutoff dates, meaning they might not know about the latest features, API changes, or best practices for our libraries.

Our MCP server solves this by providing:

  • Fresh Documentation: Real-time access to the latest version of our documentation.
  • RAG-based Solutions: A custom Retrieval-Augmented Generation (RAG) solution that finds relevant information and generates precise code snippets based on your specific queries.

When enabled, your AI agent can "ask" our server for help, ensuring that the code it generates for you is accurate, up-to-date, and follows our recommended patterns.

Server URL: https://docs.svar.dev/mcp

Hosted service​

The SVAR MCP server is provided as a fully managed, hosted service. Due to its high resource requirements and stateless nature (it does not access or store local user data), it is not available for local deployment.

Service doesn't store any personal information of MCP users. Queries sent to the server may be stored for debugging and service improvement purposes though. If you want to opt-out of this, please contact us at info@svar.dev, we can provide a commercial version of the service with guaranteed no-logging policies.

Integration guides​

Here are instructions on how to enable the SVAR MCP server in popular AI coding tools.

Claude Code​

To use the MCP server with Claude Code, you can add it via the command line or configure it in your project settings.

Using the CLI:

Run the following command in your terminal:

claude mcp add --transport http svar-mcp https://docs.svar.dev/mcp

Manual Configuration:

You can also add it to your ~/.claude.json or .mcp.json configuration file:

{
  "mcpServers": {
    "svar-mcp": {
      "type": "http",
      "url": "https://docs.svar.dev/mcp"
    }
  }
}

For more information, consult the official documentation.

Cursor​

Cursor has built-in support for MCP servers.

  1. Open Cursor Settings (Cmd+Shift+J or Ctrl+Shift+J).
  2. Navigate to Tools & MCP.
  3. Click Add Custom MCP.
  4. Enter server configuration
{
  "mcpServers": {
    "svar-mcp": {
      "url": "https://docs.svar.dev/mcp"
    }
  }
}

Once added, you can reference it in your chats (e.g., "Check SVAR Docs for how to configure the menu").

Alternatively, you can install the SVAR MCP server directly in Cursor with one click:

Install MCP for Cursor

Learn more in the official documentation.

VS Code​

Click the button to install the SVAR MCP server directly in VS Code:

Install MCP for VS Code

Gemini CLI​

To use the server with the Gemini CLI tool:

gemini mcp add --transport http svar-mcp https://docs.svar.dev/mcp

Alternatively, you can configure it manually:

  1. Locate your configuration file, typically at ~/.gemini/settings.json.
  2. Add the server to the mcpServers object:
{
  "mcpServers": {
    "svar-mcp": {
      "url": "https://docs.svar.dev/mcp"
    }
  }
}
  1. Restart the Gemini CLI.

See the official documentation for further details.

Antigravity (Google)​

For Google's Antigravity agent:

  1. Open command palette and type "mcp add"
  2. Select "HTTP"
  3. Enter the URL https://docs.svar.dev/mcp.
  4. Enter the Name svar-mcp.

Refer to the official documentation for more guidance.

Other IDEs / Tools​

  1. Check the settings for Model Context Protocol or Context Sources.
  2. Add a new custom source with the URL https://docs.svar.dev/mcp.

ChatGPT​

To enable the server in ChatGPT:

  1. Open ChatGPT and navigate to Settings / Apps & Connectors.
  2. Click Advanced settings.
  3. Enable the Developer mode
  4. Go "Back" and now you must have the "Create" button on the "Connectors" screen, click it
  5. Enter Name svar-mcp
  6. Enter URL https://docs.svar.dev/mcp
  7. Select No authentication as Authentication
  8. Click Create to finish

You can now ask your Custom GPT to "Consult the SVAR MCP server" when asking questions about Svelte Core.

Full setup details are available in the official documentation.