AI

Dashdown is built for AI in two directions — the LLM that helps your readers understand a dashboard, and the coding agent that helps you build one.

AI in your dashboard — <Ask /> #

Most dashboards stop at the chart. <Ask /> goes one step further: it sends a query's result to an LLM and renders the natural-language read-out right beside it — "downloads are up 12% month-over-month, driven by the pip channel." One tag turns a table into an explained insight, cached so repeat views don't re-bill, and pointed at any provider (Mistral, Claude, OpenAI, OpenRouter).

<Ask data={downloads_by_month} ask="Summarize the download trend in two sentences." />

→ Ask — LLM commentary for the full component, semantic-layer support, providers, caching, cost and the safety model.

AI that builds your dashboard — coding agents #

Because a dashboard is just Markdown + SQL + component tags, a coding agent (Claude Code, Cursor, Codex, …) can author it directly. Every project ships a tool-agnostic guide, and the CLI exposes the framework's own knowledge so the agent checks facts instead of guessing.

→ Coding agents for AGENTS.md, dashdown skill, the CLI loop, and llms.txt.