Connect AI

Need desktop Hub editing over MCP (open charts in Chartbuddy Hub, list_charts, …)? See Chartbuddy Hub – Connect MCP.

How Embed fits AI workflows

1. Fetch llms.txt (curated index).
2. Fetch only the docs pages you need (quick start, chartData schema, chart type, axes, validation) via .md links from the index — do not load llms-full.txt by default.
3. Write HTML/JS that imports @chartbuddy.io/embed and passes chartData.
4. Size the host – definite width/height (aspect-ratio or fixed px). See Canvas & sizing.
5. Verify with Visual QA – layout first, then annotations.

Pick your client

Client	Guide
Cursor	Connect Cursor
Claude	Connect Claude
Claude Code	Connect Claude Code
OpenCode	Connect OpenCode
ChatGPT / Codex	Connect ChatGPT / Codex

Shared handoff

Paste into a new agent chat when starting an Embed task:

Build ChartBuddy Embed charts with @chartbuddy.io/embed (new Insight()).
1) Fetch https://unpkg.com/@chartbuddy.io/embed/llms.txt (index)
2) Fetch linked docs .md pages you need (quick-start, chart-data, chart type, validation) — do not load llms-full.txt by default
Host sizing: mount element needs definite width/height (prefer width + aspect-ratio, not height:100% in flex/slide layouts).
Use validateChartData / Visual QA loop before declaring done.

Troubleshooting

Symptom	Check
Charts render tiny or letterboxed	Host sizing – aspect-ratio or fixed px
format changes do nothing	Set numberFormat: { mode: 'custom', customFormat } – see Number formats
Agent invents wrong API	new Insight() only – not Chart.js / createChart
Slide deck charts wrong size	Mount when slide is visible; remount or update after layout
Need Hub window tools	Use Hub Connect MCP instead of Embed docs alone