Chat with your Postgres databaseand trust the number
Ask your Postgres database a question in plain English and get a chart back with the verified, read-only SQL underneath it — Chion never improvises a query against your production data, it answers only from the SQL your team already trusts, and the whole skill set exports to Claude Code, Cursor, or Codex.
TWO WAYS TO RUN
Insights or any chatbot
Use the skills natively inside Chion's Insights Generator, or export them via MCP to Claude, ChatGPT, Gemini, or Cursor.
Your team chats with Postgres; every answer traces back to verified SQL
AskYourDatabase runs CRUD on your tables; DIY GPT-4o tutorials improvise SQL. Chion connects the SQL your team already trusts and answers read-only — every chart traces back to a verified query.
The broken analytics loop
Stale dashboards. Disconnected AI. Blocked analysts. Chion closes the loop, from question to chart in seconds.
Chion compiles your team's verified queries into a portable skill library. Runs in Chion Studio or any LLM, with the same audit trail, the same row-level security, and every chart traceable to exact SQL.
Text-to-SQL tools hallucinate queries against your live database
Staledashboards.
DisconnectedAI.
Blockedanalysts.
Chion is your verified analytics team — finance, ops, growth agents, scoped by persona. From question to chart in seconds. Drop your agent + skills file into Claude Code, Codex, or Cursor — or chat with it inside Chion, where every answer routes to a verified query.
Your verified queries already exist.
One bulk-upload. Chion ships you a per-persona CHION.md agent. Domain-aware, table-scoped, ready for Claude Code, Codex, or Cursor.
No raw rows sent to the LLM.
The model only sees controlled summaries, counts, samples, and aggregations, not direct table access.
Read-only and encrypted by default.
SELECT-only credentials. AES-256 vault. Your data never moves.
Traceable SQL at every step.
Every answer shows the exact query that generated it.
Zero migration. Zero pipelines.
Connect to your database and query it where it lives.
Deploy on your terms.
Managed cloud today; dedicated infrastructure on the Enterprise roadmap. Same pipeline, same guarantees, wherever it runs.
Under the hood.
A multi-phase AI pipeline, not a wrapper around another LLM.
See the 13-step verified SQL pipeline from schema profiling to grounded narrative.
Chion vs. text-to-sql chat tools: verified and read-only, not improvised
Chion ships a verified, per-persona agent. Per-seat tools share one workspace. Notebooks regenerate SQL every run. Here is how they compare.
| Chion | Per-seat AI analytics | Notebook-based AI analytics | |
|---|---|---|---|
| Verified-only answers (no hallucinated SQL) | Routes to verified query, deterministic execution | Runs LLM each turn; schema hallucination risk | Runs LLM each turn; schema hallucination risk |
| Bulk-upload your verified queries, Day 1 | Bulk-upload or one-by-one; auto-organized by persona | Manual semantic-layer setup | Multi-month modeling |
| Leaders chat with persona-agent directly | Finance / ops / growth agents chat-ready | Analyst-led workflow | Notebook-led workflow |
| Portable AI agent file: CHION.md → Claude Code, Codex, Cursor | Portable agent file you own; drops into any IDE | Locked to vendor UI | Locked to vendor UI |
| Per-persona scoping (RLS-governed) | Different agents for finance, ops, growth | One workspace per team | One workspace per team |
| Setup model | One-time bulk-upload, compile, run | Recurring per-seat license + ongoing config | Recurring per-seat license + multi-month modeling |
Full comparisons: Chion vs BI dashboards · Chion vs other text-to-SQL tools.
Read-only SELECT, row-capped, RLS-enforced — the LLM physically cannot mutate your data
We connect with the read-only role you give us. Your data stays in your database. We send schema metadata to the model, run the query, and render the chart. That's the entire data flow.
Encrypted credential vault
Credentials encrypted at rest. Decrypted in-memory only. Never logged.
Read-only by default
SELECT-only permissions. We will never modify or delete your data.
Row-level security honored
Your RLS policies are honored. Every query runs with your connection role's permissions.
Same pipeline everywhere: managed cloud, CoreWeave, or on-premise.
How it works
Four steps from a question to your team's portable AI brain.
Connect Your Data
Point us at your PostgreSQL database. Two minutes, read-only credentials, done.
Semantic Layer
Automatic Discovery
We scan every table, column, and relationship the moment you connect. No manual setup; the system figures out your schema so you don't have to explain it.
Ask a question about your data…
Ask Your Data
Ask a question, get a chart. The SQL behind every answer is real, verified, and visible. Nothing fabricated, nothing hidden.
Generate your real file from Studio. Every verified query becomes a skill. Learn more →
A Skill for Every Question
Every verified query becomes a reusable skill, saved to a personal CHION.md file: one per employee, governed by your database permissions.
Export the bundle. Claude, Codex, and Cursor inherit your team's analytics know-how.
Built by an analyst who got tired of waiting.
Founded in 2024 by Jonathan Dag, former Meta, Twilio, and American Express analyst.Read allJonathan Dag, Founder
After a decade building data infrastructure across Meta, Twilio, American Express, MasterCard, and Bosch, the pattern was always the same: analysts buried in SQL, business users waiting days for a dashboard update, and AI tools hallucinating the answer.
Chion is the analytics layer that should have existed from the start. Ask in plain English, get a verified answer backed by real, traceable SQL.
That's the tool I kept wanting someone else to build.
Choose how you deploy your portable SQL skills.
Quick Start
Connect PostgreSQL. Ask a question. Get a chart with the SQL behind it. Takes about two minutes.
- Fast models for speed, advanced models for depth, selected dynamically
- Search across your schema by meaning, not just column names
- Connect your PostgreSQL database
- Skip the ticket queue
Custom Setup
Pick your own model. Deploy on CoreWeave or your hardware. Same pipeline, same verification, just running where you control it.
- PostgreSQL: the only language you need
- Request your own model & infrastructure
- Anthropic, OpenAI, Gemini, Mistral
- On-prem or dedicated GPU options
- CoreWeave cloud infrastructure
Agent Skills
Every verified query becomes a CHION.md skill: a portable SQL Analytics Agent file interchangeable with CLAUDE.md, AGENTS.md, and SKILL.md. Natively compatible with Claude Code and Codex.
- One Markdown file, four idiomatic mirrors
- Auto-promoted from your real queries
- Per-domain, per-table, per-employee scoping
Frequently asked questions
Browse 145 answers about security, pricing, pipeline architecture, and integrations. Everything from how SQL is verified to how credits and plans work.
Explore Chion
Solutions
Learn
Last reviewed: May 26, 2026







