Agent-first
An MCP server your agent already knows how to use. You ask in plain language; your agent handles the rest.
Semantic analytics
Multimodal analytics for AI agents — with a budget.
SQL on Unstructured Data for Agentic Exploration.
Agent chat
Why SundaeDB
Built for agents, not dashboards.
Multimodal
Analyze CSV and ZIP files with text, images, or audio — ask questions across mixed data in one workflow.
Cost controls
Your agent sets a per-query budget ceiling. Pay for actual usage via x402 — no accounts, no subscriptions.
Partial results
Set a tight budget and SundaeDB returns the best answer it can within that cap. When the money or time runs out, you still get a useful approximation.
How it works
Three steps to get started.
-
Step 1
Connect
Paste the connect prompt into your agent; it sets up SundaeDB and payments.
-
Step 2
Ask
Point your agent at CSV or ZIP data and ask questions in plain language.
-
Step 3
Get answers
Your agent loads data, runs analysis, and reports results back to you.
Connect
Add SundaeDB to your agent.
Paste the prompt below into Cursor, Claude Code, or any MCP-capable agent. It reads sundaedb.com/SKILL.md and configures SundaeDB plus x402 payments.
Add SundaeDB to my agent: sundaedb.com/SKILL.mdNode.js is required on the machine where your agent runs. Restart the client after setup. Paid tools are priced on the pricing page.
Pricing
Pay per query. Nothing else.
No accounts and no subscriptions. Your agent pays for each call directly over x402. A tight budget may yield a rough approximation; your agent can raise the cap for a fuller answer.
Queries
from 1¢
Your agent sets the cost bound per query. Settlement is for actual usage, capped at the authorized maximum.
Sandbox database
5¢
Instant throwaway database for one-off analyses. No setup required.
Everything else
Free
Schema inspection, table previews, uploads, and documentation.
Prices are tax-inclusive. Payments are non-refundable, including when a query returns partial results within its budget cap.