Just one more query.
Columnbo is conversational analytics for people who'd rather ask a question than build a dashboard. Real answers from real data — no hallucinations, no SQL, no waiting on the data team.
Columnbo is conversational analytics for people who'd rather ask a question than build a dashboard. Real answers from real data — no hallucinations, no SQL, no waiting on the data team.
There's just one more thing that's been bothering me. See, most analytics tools — they give you charts. But they never tell you what the chart means. Columnbo does.
In plain English. "How many customers churned last month?" "What's our best-selling SKU in Texas?" Just type it.
An LLM translates your question into a precise, validated query against your actual schema. No guessing. No hallucinated column names.
The query runs against your database. You get the insight, a chart if it helps, and a headline you could paste into Slack right now.
Other tools show you data. Columnbo tells you what it means.
Every answer traces back to a real query that ran against real data. Columnbo never makes up numbers. If the query fails, you see the failure — not a confident-sounding lie.
"How many dog cards are there?" → "How many of them are red?" Columnbo remembers context across turns, so follow-ups just work. No re-explaining what you're looking at.
Columnbo logs how people actually talk about your domain and mines those logs for vocabulary gaps, missing synonyms, and intent patterns. The system gets sharper the more people use it — not through retraining, but through structured catalog enrichment.
When a chart actually helps, Columnbo renders one — annotated, titled with the finding, direct-labeled. When it doesn't help, you get a sharp sentence instead. No chartjunk.
An LLM Guard sidecar scans every inbound question for injection attempts, PII leaks, and jailbreaks before the query pipeline ever sees it. Fail-closed in production.
Columnbo talks to any data source that speaks SQL or Malloy — DuckDB, Postgres, BigQuery, parquet files on disk. Swap the dataset; the conversation layer stays the same.
The live demo runs against 109,000+ Magic: The Gathering card printings. Ask it anything — rarity distributions, mana curves, price trends, format breakdowns.
"Do rares really outnumber commons?"
Try the demoWhether it's a sales database, a product catalog, or a proprietary dataset your analysts keep asking the same three questions about — Columnbo can sit on top of it. Let's talk about what that looks like for your team.
[email protected] →Malloy is an open-source semantic modeling language created by Lloyd Tabb, the founder of Looker. It replaces hand-written SQL with composable, reusable data models that compile to optimized queries. Columnbo uses Malloy as its query backbone — every answer traces back to a real Malloy query, not a hallucinated SQL string.
Learn more about Malloy →