The AI SQL agent that shows its work · analyst-founded
The verified-SQL analytics tool, built by the analyst who got tired of AI guessing
Chion is an AI SQL agent that answers your questions in plain English and shows the exact, read-only SQL behind every chart, so you can verify the number before you trust it. Built by a technical analyst who spent ten years at Meta, Twilio, and American Express.
10+
Years in analytics
Jonathan Dag has spent over 10 years as a technical analyst working for Meta, Twilio, American Express, MasterCard, and Bosch. Different industries, same broken loop: answers no one could verify. Business users waited days for a dashboard update. Analysts rewrote the same SQL with different filters. AI tools made up queries that looked right. For a decade that loop was merely slow. Now that teams are handing analytics to AI agents, an answer you can't verify or take with you is a real liability. Chion exists to close that gap.
Ask in plain English. Get verified SQL. See the chart. Trust the answer.
What is Chion?
So he built it. Chion is the AI SQL workforce: verified SQL agents that export as SKILL.md files and run in any AI tool, whether that's Claude, Cursor, Codex, or your own. The analytics knowledge stays yours instead of being locked to us or to a single model. Each agent runs read-only against PostgreSQL, with credentials in an AES-256-GCM vault, every query logged to an immutable audit trail, and the SQL shown beside every chart.
We won't ship a chart you can't trace back to its SQL
Why we exist, and what we won't compromise.
If you can't see the SQL, you can't trust the chart. If you can't trace the number, you shouldn't put it in a slide deck. We built Chion on three principles we won't trade for speed:
Show the work
Every chart ships with the exact SQL that produced it. No exceptions.
Built for analysts, usable by everyone
The pipeline validates like an analyst would. The interface doesn't require you to be one.
Trust by default
SELECT-only. Read-only. Raw row-level data never reaches our LLM providers. You choose where every component runs: our cloud or your own.
Show the work. That's the whole idea.
How every answer gets verified
Show the work at every step, for every request.
13+
Workflow steps per request
Every question runs the full verification framework: schema profiling, two-layer validation, and repair.
7+
Analytic strategies
Each request routes across seven analytical strategies, from entity lookup to extrema detection.
<30k
Tokens per request
However complex the question, the whole pipeline stays under 30k tokens. Fast, and predictable on cost.
None of this is a settings toggle. It's how the pipeline is built: read-only on your Postgres, portable to any AI tool.
Chion enforces SELECT-only SQL, vault-encrypted credentials (AES-256-GCM), row-limit caps (1,000 rows max), an immutable audit log of every query, and full tenant isolation. Raw row-level data never reaches our LLM providers; only metadata and aggregated query results are transmitted.
Talk to the founder
Jonathan reads every email. Send questions, feedback, or a demo request.
Jonathan Dag · Founder & CEO, Chion
A former technical analyst at Meta, Twilio, American Express, MasterCard, and Bosch. Engineer by training, analyst by trade.
LinkedIn
X
YouTube
InstagramWhere to find Chion
External profiles, launches, and review platforms, all in one place.
Chion: AI SQL workforce for PostgreSQL
Chion Studio reviews and ratings on G2
Chion company profile on Crunchbase
Chion is a solo-founder operation today. If you're an engineer or analyst who wants to help build analytics that teams can defend, I'd like to hear from you. Email contact@chion.ai.
Frequently asked questions
See the work behind every number.
7-day trial. Connect a read-only PostgreSQL database and ask your first question.
Last reviewed: June 9, 2026