Plotly does not publish an MCP server, and the plotly/mcp-server repository this page used to link does not exist. Several community servers wrap Plotly, and they fall into two groups: one renders the chart inside the conversation as an MCP App, the rest write HTML or PNG files that you open yourself.
Source and setup verified on 2026-11-03
Visualizationjenreh · updated 2026-11-02
FastMCP server on Plotly 6. Each chart tool returns a chart_view MCP App view, so the chart appears in the conversation with zoom, pan and PNG export.
Best for: Interactive charts inside Claude or another MCP Apps host.
a-humphrey · ★ 1 · updated 2026-03-03
Python server covering all 51 Plotly trace types, multi-panel dashboards, and charts straight from CSV, Excel or JSON files. Saves HTML, PNG or both.
Best for: Complex charts and dashboards saved as files.
StarRocks (official) · ★ 196 · updated 2026-09-08
Official StarRocks server. Its query_and_plotly_chart tool runs SQL and writes the result as a Plotly HTML chart.
Best for: Charting data that already lives in StarRocks.
arshlibruh · ★ 9
Built for Cursor. Early stage: 5 of 49 planned trace types (scatter, bar, line, pie, histogram).
Best for: Basic charts in Cursor.
{
"mcpServers": {
"plotly-charts": {
"command": "/path/to/plotly_mcp/.venv/bin/plotly-mcp"
}
}
}Create the virtualenv and run pip install -e . inside the cloned repository first.
claude mcp add plotly-charts /path/to/plotly_mcp/.venv/bin/plotly-mcpNo. Plotly has not released one, and there is no plotly/mcp-server repository.
appkit-mcp-charts. It returns an MCP App view, which Claude and other MCP Apps hosts render inline. The others save files.
plotly_mcp by a-humphrey passes any Plotly trace or layout property through, covering all 51 trace types.
Interactive 3D globe viewer using CesiumJS for geospatial visualization
Interactive 3D scene renderer with Three.js for stunning visualizations
Real-time GLSL shader renderer for creative visual effects