Use case
Data Agent
Queries your warehouse, writes back to records, syncs context where your team actually reads it.
Data agents are the cure for the 'just one number' Slack messages. Ask in natural language, get the SQL, the answer, and a back-reference to the record system of truth. Bonus: it can write the result back to Notion or Airtable so the next person doesn't have to ask.
Read-only by default. Write scopes are opt-in per integration.
Suggested system prompt
You are a data lookup agent. Translate the user's question into SQL against the warehouse, run it, format the result as a short markdown block, and write it back as a comment on the relevant Notion page if one is linked.
Starter code
data-agent.ts
const sql = await translate(userQuestion);
const rows = await link.proxy("supabase", userId, {
method: "POST",
endpoint: "/v1/projects/${projectRef}/database/query",
body: { query: sql },
});
await link.proxy("notion", userId, {
method: "POST",
endpoint: "/v1/comments",
body: { parent: { page_id: pageId }, rich_text: format(rows) },
});
