connect a database, ask a question. ryuk reads your schema, writes the SQL, runs it, and shows its work — so you can verify and move on. no copy-pasting schemas into chatgpt.
SELECT region, SUM(revenue) AS total FROM sales_metrics WHERE quarter = 'Q4' GROUP BY region
paste a postgres, mysql, or warehouse connection string. read-only by default. credentials encrypted, never logged.
tables, columns, types, and relationships indexed on connect. ryuk knows which user_id joins where before you ask.
plain questions. ryuk drafts the SQL against your actual schema, validates it, and runs it. no prompt engineering.
every answer comes with the SQL and the reasoning that produced it. accept, tweak, or copy it out to your own editor.
save a query as a metric. arrange metrics into dashboards that refresh on load. no chart library to learn.
turn any conversation or dashboard into a clean PDF. share it without a login — or stop sending screenshots in slack.
ryuk reads your actual tables before answering. the SQL references columns that exist, joins that are correct, and types that match. every conversation is saved — revisit the query, export the PDF, or share the result without pasting screenshots in slack.
joining sales_metrics on region and quarter. filtering for Q4 2025 and Q4 2024. aggregating revenue with YoY delta.
yes, but connections are read-only by default. ryuk can't write, update, or drop unless you explicitly flip that off — and most people never do. if you're still nervous, point it at a read replica.
you see every query before it runs. reject it, edit it inline, or accept it. ryuk also validates its own output against your indexed schema before execution, so hallucinated columns and bad joins get caught before they hit your db.
encrypted before they hit disk. never written to logs. never sent to the LLM in a prompt — only table and column names are. the credential itself stays on our infrastructure.
postgres and mysql are first-class. bigquery, snowflake, and redshift work through their standard drivers. mongo, dynamodb, and other non-SQL stores aren't supported and probably won't be — ryuk is built around SQL.
no. queries go to the configured LLM provider (anthropic or openai) under their standard API terms. no training opt-in happens on our side, and result data never leaves the request/response cycle.
everything stays. you drop to the free tier, keep your connections, metrics, and reports, and just get fewer credits per month. no lock-in, no export fee, no deletion threats.
still have a question? [email protected]
1 credit = 1,000 AI tokens. start free. top-ups never expire. cancel in two clicks.
the whole thing, slower. enough to try anything.
if you open ryuk most days.
if this is running against real data every day.
all plans include every feature. see full pricing details
free to start. first connection takes about a minute. no credit card.