mcp-highcharts
# mcp-highcharts
Render interactive [Highcharts](https://www.highcharts.com/) charts inline in AI chat — VS Code, GitHub Copilot, Claude Desktop, or any MCP client with [MCP Apps](https://modelcontextprotocol.io/extensions/apps/overview) support.
[](https://insiders.vscode.dev/redirect/mcp/install?name=mcp-highcharts&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22mcp-highcharts%40latest%22%5D%7D)
[](https://insiders.vscode.dev/redirect/mcp/install?name=mcp-highcharts&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22mcp-highcharts%40latest%22%5D%7D&quality=insider)
[](https://www.npmjs.com/package/mcp-highcharts)

## Setup
Click a badge above, or add to your MCP config:
```json
{
"mcp": {
"servers": {
"highcharts": {
"command": "npx",
"args": ["-y", "mcp-highcharts@latest"]
}
}
}
}
```
<details>
<summary>Claude Desktop</summary>
Add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"highcharts": {
"command": "npx",
"args": ["-y", "mcp-highcharts@latest"]
}
}
}
```
</details>
<details>
<summary>HTTP mode</summary>
```bash
npx mcp-highcharts@latest --http # http://localhost:3001/mcp
PORT=8080 npx mcp-highcharts@latest --http # custom port
```
</details>
## Tools
| Tool | Description |
|------|-------------|
| `render_chart` | Any chart type — line, bar, pie, scatter, heatmap, sankey, gauge, treemap, wordcloud, and [60+ more](https://www.highcharts.com/demo) |
| `render_stock_chart` | Financial charts with navigator, range selector, and 40+ technical indicators |
| `render_dashboard` | Multi-component layouts with charts, KPIs, and data grids |
| `render_map` | Choropleth maps, map bubbles, map points (auto-fetches map data from CDN) |
| `render_gantt` | Project timelines with tasks, dependencies, and milestones |
| `render_grid` | Standalone data tables with sorting, pagination, and formatting |
All tools accept the full [Highcharts Options API](https://api.highcharts.com/highcharts/).
## Prompts
| Prompt | What it does |
|--------|-------------|
| `chart_from_data` | Paste data → get the best chart |
| `dashboard_layout` | Scaffold a dashboard |
| `stock_analysis` | Candlestick + volume + indicators |
| `comparison_chart` | Side-by-side comparison patterns |
| `project_timeline` | Gantt with dependencies |
| `live_chart` | Live-updating chart from a data URL |
## Configuration
### Theming
Charts auto-adapt to host light/dark mode. Override with environment variables:
```json
{
"env": {
"HIGHCHARTS_THEME": "dark-unica",
"HIGHCHARTS_OPTIONS": "./my-theme.json"
}
}
```
`HIGHCHARTS_OPTIONS` accepts `.json`, `.js`, `.mjs`, `.ts`, or inline JSON.
**Built-in themes:** `adaptive` (default), `avocado`, `brand-dark`, `brand-light`, `dark-blue`, `dark-green`, `dark-unica`, `gray`, `grid`, `grid-light`, `high-contrast-dark`, `high-contrast-light`, `sand-signika`, `skies`, `sunset`
### Schema depth
Controls how much type information is sent to the LLM:
```json
{ "env": { "SCHEMA_DEPTH": "1" } }
```
| Depth | Description |
|-------|-------------|
| `0` | Property names only — zero context overhead |
| `1` (default) | Top-level types + descriptions + examples |
| `2` | One level of typed children |
| `3` | Two levels deep — complete recursive Highcharts type tree |
### Data sources
For live-updating charts, use the Highcharts data module with `data.csvURL` and `data.enablePolling: true`.
### Image export (non-app fallback)
For MCP clients that don't support [MCP Apps](https://modelcontextprotocol.io/extensions/apps/overview), enable server-side image export to include a PNG screenshot of each chart in the tool response:
```json
{
"env": {
"IMAGE_EXPORT": "true"
}
}
```
Charts are rendered as PNG and returned as base64 image content blocks. The interactive MCP app is still included for capable clients.
**Rendering strategy (automatic):**
1. **Local (Puppeteer)** — if [`highcharts-export-server`](https://github.com/niclasvaneyk/highcharts-export-server) is installed, charts render locally via a headless browser. No network calls, fastest option.
2. **Remote fallback** — if local isn't available, chart config is POSTed to the [Highcharts Export Server](https://export.highcharts.com/).
To enable local rendering, install the optional peer dependency:
```bash
npm install highcharts-export-server
```
To use a custom remote export server:
```json
{
"env": {
"IMAGE_EXPORT": "true",
"EXPORT_SERVER_URL": "https://your-export-server.example.com/"
}
}
```
You can also enable it programmatically: `createServer({ imageExport: true })`.
> **Note:** Image export works for standard charts, stock charts, and Gantt charts. Dashboards, data grids, and maps with string-based map keys (e.g. `"custom/world"`) are not exportable and will return text-only results.
## Development
```bash
npm install
node scripts/generate-from-tree.mjs --multi # generate Zod schemas at depths 0, 1, 2
npm run build
npm test
```
### Project structure
```
main.ts Entry point (stdio + HTTP transports)
server.ts MCP server — tool registrations and handlers
src/
export-image.ts Server-side PNG export (local Puppeteer + remote fallback)
input-schema.ts Depth-based schema selection + LLM-friendly overrides
mcp-app.ts Client-side Highcharts rendering
module-loader.ts Dynamic Highcharts module loading
generated/ Auto-generated from Highcharts API (do not edit)
highcharts-depth-{0,1,2}.gen.ts Zod schemas at each depth
module-map.json Chart type → Highcharts module mapping
scripts/
generate-from-tree.mjs Generate Zod schemas from Highcharts tree.json
generate-module-map.mjs Generate module-map.json from Highcharts
example-providers.mjs Example extraction for schema generation
measure-schema.ts Measure tool context size at each depth
```
Chart types, module mappings, and schemas are auto-generated from the installed Highcharts version — just `npm update highcharts` and regenerate.
## License
MIT
TDQS
Scored across 6 tools
Each tool has a clearly distinct purpose targeting specific Highcharts chart types or components (e.g., standard chart, dashboard, Gantt, grid, map, stock chart). There is no overlap in functionality, and the descriptions clearly differentiate each tool's use case, preventing misselection.
All tool names follow a consistent verb_noun pattern with 'render_' as the prefix, followed by the specific chart or component type (e.g., render_chart, render_dashboard). This predictable naming scheme enhances readability and usability across the tool set.
With 6 tools, the server is well-scoped for rendering various Highcharts visualizations, covering key chart types and components. Each tool earns its place by addressing distinct visualization needs, making the count appropriate for the domain without being overwhelming or insufficient.
The tool set provides complete coverage for rendering Highcharts visualizations, including standard charts, dashboards, Gantt charts, data grids, maps, and stock charts. This covers the core Highcharts library and major extensions, leaving no obvious gaps for the server's purpose of chart rendering.