Skip to main content

MCP Setup

Use Chart-Output from Claude.ai, ChatGPT, Claude Code, Cursor, or Windsurf. Ask your AI to generate a chart and it renders inline — no code required.

There are two ways in: the hosted server at https://www.chart-output.com/mcp, which works in chat apps with nothing to install, and the local npm package for editors, which has more tools.

Hosted server (Claude.ai, ChatGPT)

Add this URL as a custom connector. It speaks MCP over streamable HTTP and has one tool, render_chart, which returns the chart as an image in the chat plus a link to keep editing it in the playground.

text
https://www.chart-output.com/mcp
  • Claude.ai: Settings → Connectors → Add custom connector, then paste the URL.
  • ChatGPT: add it as a custom connector from Settings. Depending on your plan, ChatGPT may ask you to turn on developer mode first.
  • Claude Code:
bash
claude mcp add --transport http chart-output https://www.chart-output.com/mcp

No key, or your own key

Without an API key the hosted server renders PNGs up to 1200px a side, with no custom logos or saved brand kits, from a free daily budget shared by everyone using it. When that runs out, the tool says so. Add your own key to render on your plan's quota, sizes and brand kits:

  • Clients that can send headers (Claude Code, Cursor): Authorization: Bearer pk_live_…
bash
claude mcp add --transport http chart-output https://www.chart-output.com/mcp \ --header "Authorization: Bearer pk_live_YOUR_KEY"
  • Chat apps that only take a URL (Claude.ai, ChatGPT): add the key to the URL, https://www.chart-output.com/mcp?key=pk_live_YOUR_KEY. Anyone who sees that URL can render on your quota, so don't share it, and revoke the key if it leaks.

Renders with a key count against your plan exactly like POST /api/v1/render, and the hosted server always returns PNG.

Local server prerequisites

The @chartoutput/mcp package runs on your machine over stdio. Besides render_chart it has render_card, render_chart_ai, brand kit tools, and gallery and example lookups.

Configure your client

The JSON config is the same across clients; only the file path changes (and VS Code uses a "servers" key). Pick your editor below, paste the block, and swap in your API key.

// .mcp.json { "mcpServers": { "chart-output": { "command": "npx", "args": ["@chartoutput/mcp"], "env": { "CHART_OUTPUT_API_KEY": "pk_test_YOUR_KEY" } } } }

Claude Code: Create this file in your project root. Replace `pk_test_YOUR_KEY` with your key from the dashboard.

Verify it's working

Restart your editor so it picks up the new MCP server, then ask your AI:

text
Create a bar chart showing Q1 through Q4 revenue: 120k, 145k, 132k, 178k

You should see a chart image rendered inline in the chat. If you get a 401 error, your API key is missing or invalid — double-check the CHART_OUTPUT_API_KEY value in your config file.

What you can ask

Natural language is enough — the MCP server handles chart specs, API calls, and image delivery.

Ask

“Create a bar chart showing Q1 through Q4 revenue: 120k, 145k, 132k, 178k”

You get

Inline bar chart PNG rendered via `render_chart`.

Ask

“Make a line chart of monthly active users from 12k in January to 28k in December”

You get

Inline line chart with labeled axes and a single dataset.

Ask

“Render a dashboard card for API latency with a dark theme, KPI strip, and footer”

You get

Full branded card image via `render_card` — header, KPI row, chart, and footer in one PNG.

Ask

“Chart this CSV of signups by week and pick the best chart type”

You get

AI-generated chart via `render_chart_ai` (Pro or Agency API key required).

Package & source

Related walkthroughs