What the MCP server actually does

When you ask your coding agent to build a chart, it calls get_lightningchart_context from the MCP server. The server returns the official LightningChart JS agent instructions: the correct package name, current best practices, common mistakes to avoid, and pointers to the right documentation. The agent uses this context before generating any code.

In agents that show tool activity, you will see the get_lightningchart_context call happen automatically when the agent decides it needs charting guidance. In agents that do not, adding one sentence to your prompt is enough:

Setup: Claude Code

The fastest method is the CLI one-liner:

claude mcp add lightningchart-js -- npx -y @lightningchart/mcp-server@latest

That is it. Restart Claude Code and open a new session. The server is active.

If you prefer to add it through the config file instead, open ~/.claude/config.json (or the Claude Code settings UI) and add:

{
  "mcpServers": {
    "lightningchart": {
      "command": "npx",
      "args": ["-y", "@lightningchart/mcp-server"]
    }
  }
}

Restart Claude Code after editing the config. Check the MCP server list in settings to confirm lightningchart appears with a green status indicator.

Setup: Cursor

Open Cursor settings and navigate to the MCP section. Click “Add MCP server” and fill in:

Save and restart Cursor. In a new chat, you should see lightningchart listed under available tools.

If Cursor uses a JSON config file in your project (.cursor/mcp.json) rather than the UI, use the same JSON block as the Claude Code config above.

Setup: Cline, Roo Code, Windsurf, and others

Any MCP-compatible agent uses the same config. Find the MCP server configuration in your tool’s settings and add:

{
  "mcpServers": {
    "lightningchart": {
      "command": "npx",
      "args": ["-y", "@lightningchart/mcp-server"]
    }
  }
}

The npx command downloads and runs @lightningchart/mcp-server from npm automatically. You do not need to install it globally.

Verifying it works

Start a new session and ask for a chart:

Ask a prompt that clearly involves charting / data visualization. This can be in plan mode if you don’t want any actual file changes immediately.

Practically all models show progress information and tool activity before printing the final output. During this crunching stage, expand the details to see how they source information for LCJS:

• The agent calls get_lightningchart_context before generating any code (visible in tool call logs).
• It uses @lightningchart/lcjs as the import, not @arction/LCJS (the old package name).
• It asks about or correctly wires a license key.
• It does not mix in Chart.js or Recharts syntax.

For any charting-related task, the agent will pull up the “lightningchart” MCP server for context before starting.

Troubleshooting

The most common issue is the MCP server not appearing after adding the config. Check these in order:

• Did you restart or reload the client? MCP servers are loaded at application startup.
• Is the config in the right file? Claude Code uses ~/.claude/config.json. Cursor uses its own config path.
• Is npx available? Run npx --version in your terminal to confirm Node.js is installed.
• Is it visible in settings? Does the tool list show a lightningchart server? If not, the config was not picked up.

The GitHub repo README at github.com/Lightning-Chart/lightningchart-js-mcp-server has a dedicated troubleshooting section covering edge cases for specific clients.

What to try once it is running

A few prompts that work exceptionally well with the MCP server active:

The agent will call get_lightningchart_context on any prompt where it decides LCJS guidance is relevant, which in practice means any prompt that mentions charting, dashboards, or data visualization in a JavaScript or TypeScript project.