FOR DATA SCIENTISTS

Explore JSON like
it's a dataframe.

Group, filter, aggregate, and reshape messy nested API responses — by asking. QDev writes the jq, runs it locally, and hands you a clean table to drop into pandas or polars.

✓Aggregations are exact — jq does the math, not the model
✓The jq expression isthe artifact — reproducible & shareable
✓Sensitive datasets stay on your machine — never uploaded

Add to Chrome Add to Firefox

⌘K

try

● TRANSACTIONS.JSON — 4,812 ROWS · LOCAL

{
  "transactions": [
    {
      "id": "t_5012",
      "user": "ada",
      "category": "compute",
      "amount": 420.5,
      "date": "2026-01-14"
    },
    {
      "id": "t_5013",
      "user": "ben",
      "category": "storage",
      "amount": 88.2,
      "date": "2026-01-15"
    },
    {
      "id": "t_5014",
      "user": "ada",
      "category": "compute",
      "amount": 512,
      "date": "2026-02-02"
    }
  ]
}

↑ Ask a group-by or total — jq runs on all 4,812 rows locally. Only the schema is ever sent.

exact totals · reproducible jq · nothing uploaded

SOUND FAMILIAR?

The JSON-wrangling tax

The jq you always re-Google

group_by then map then add — you know the shape, never the exact incantation. Twenty minutes and three Stack Overflow tabs later.

Nested arrays that won't flatten

API gives you users[].transactions[] and your notebook chokes. json_normalize only gets you halfway.

Files too big to eyeball

A 40 MB export won't open in the editor, and you just need one grouped total to sanity-check the pipeline.

ASK, DON'T GOOGLE THE JQ

Every wrangle, in plain English

📊

Group & aggregate

Sums, counts, averages by any field — computed by jq, so the numbers are exact.

average order value by region

🧹

Flatten nested arrays

Explode users[].posts[] into flat records ready for a dataframe.

flatten every post into one array with the user's name

🔍

Filter & slice

Pull just the rows you want, then keep only the columns that matter.

failed payments in the last 7 days

🔀

Reshape & pivot

Pivot long to wide, rename keys, coerce shapes — no bespoke script.

pivot status counts into one row per day

FITS YOUR NOTEBOOK

From raw JSON to a dataframe in four steps

Paste or open JSON

Any API response, log dump, or export — even multi-MB.

Ask in words

⌘K → “total spend by category.” QDev writes the jq.

Get a clean table

Exact result, rendered tabular — copy as CSV or records.

Into pandas / polars

Paste records straight into a DataFrame and keep going.

# QDev: "total spend by category" → copy records import pandas as pd df = pd.DataFrame(records) # tidy, typed, ready

● BUILT FOR SENSITIVE DATA

Your dataset never leaves the tab.

PII, financials, health records — the values stay local. Only the schema (field names & types) and your question are ever sent, and only on the hosted tier. Prefer zero egress? Run local AI, free, and nothing leaves at all.

✓jq runs in a sandboxed Web Worker — no network access

✓Hosted proxy is stateless & zero-retention — logs nothing

✓Local WebLLM / Ollama mode uploads nothing, not even schema

Stop hand-writing jq at 11pm.

Install free, keep your data local, and let ⌘K reshape any JSON into a table you can actually analyze.