Open your file. Ask questions. No uploads.

Drag & drop CSV / Parquet / JSON / XLSX / Avro or click anywhere to browse

Opened with DuckDB inside your browser. No accounts, no server sees your rows.

Advanced: open from URL · save database locally

Save database locally

Files

Your table will appear here. Drop a file above — it never leaves this device.

—

Table A Column A

JOIN type

Table B Column B

X Y Type Series

Rows Values

Columns Aggregation

Source

Insights (local, deterministic — no AI needed)

Local AI (optional, BYOM)

Local AI: off (deterministic insights only).

I accept the risk: model + data may exceed tab memory and crash it.

include sample rows (small tables only, stays in this tab)

Local draft by the on-device model — may contain errors. Verify before trusting; use at your own risk.

Correlations

Distribution (Box Plots)

Where do I start?

Drop a file above, or press Try sample data.
Open Profile to see column stats and data quality.
Open Chart to visualise, Analyze for insights.
In Analyze, optionally enable local AI to draft SQL from plain words — the model only sees schema + aggregates, never raw rows.

Shortcuts

Ctrl/⌘+Enter	Run SQL
Alt+1…5	Switch tabs
Alt+click cell	Show row detail
Click correlation cell	Open as scatter chart
t	Toggle theme
?	This help
Esc	Close
?threads=1	Multithreaded engine in URL (no Parquet/JSON/XLSX extensions there)

All compute runs locally in your browser (DuckDB-WASM). How privacy works

Local Analyst · build dev

Row detail

Privacy

No accounts, no tracking cookies. Nothing to log into, nothing follows you.
On this device only: theme choice, AI consent flag, saved queries (localStorage/OPFS).
On our server: anonymous download counts only; IPs truncated, kept ≤ 24 hours.
Local AI (optional): zero bundled weights; prompts contain schema + aggregates, never raw rows.
Delete everything: clear site data for this domain — nothing of yours is on our server.

Full details: About → Privacy

Local AI — voluntary opt-in

The app ships zero model weights. Nothing downloads until you explicitly choose a model. Your data never leaves this tab: a future AI rephrase would only see schema + aggregates, never raw rows.

Default recommendation: small Apache-2.0/MIT model (~700MB–1GB). Fast, permissive license.
Gemma 3.2GB GGUF: bring-your-own — open its HuggingFace page, accept the license there, then pick the local file or paste the URL below.
File stays in this tab's memory. No upload, no telemetry.