Reading and inspecting JSON from the command line

Fetch, pretty-print and filter JSON with curl and jq, reshape it for a report, and know what a JSONC config file allows.

Fetch and format

# get it and look at it
curl -s https://api.example.com/orders | jq .

# a compact single line, useful for logs and diffs
curl -s https://api.example.com/orders | jq -c .

# the same with python, if jq is not installed
curl -s https://api.example.com/orders | python -m json.tool

# send a body and read the response
curl -s -X POST https://api.example.com/orders \
  -H 'Content-Type: application/json' \
  -d '{"sku":"A-1","quantity":2}' | jq .

# see status and body separately: -w writes to stdout after the body
curl -s -o /tmp/body.json -w '%{http_code}' https://api.example.com/orders
jq . /tmp/body.json

• -s silences curl's progress meter, which otherwise corrupts the pipe. -S keeps the error messages.
• jq . is the identity filter: it parses and re-emits, which both validates and pretty-prints.
• A parse error from jq is your first signal that the response is HTML - a redirect to a login page is the classic case.
• Pipe through jq -S to sort keys, so two responses can be compared with diff.

# what did the server actually send?
curl -sI https://api.example.com/orders | grep -i content-type

# and what does the error look like when the request is wrong?
curl -s https://api.example.com/orders/999 | jq .

Selecting and reshaping

# navigate
jq '.data.orders[0].id'
jq '.data.orders | length'
jq '.data.orders[] | .id'                # one per line
jq -r '.data.orders[] | .id'             # raw strings, no quotes

# filter
jq '.data.orders[] | select(.status == "paid")'
jq '.data.orders[] | select(.total > 1000 and .currency == "GBP")'
jq '[.data.orders[] | select(.status != "cancelled")] | length'

# shape the output
jq '.data.orders[] | {id, total, currency}'
jq '{count: (.data.orders | length), total: ([.data.orders[].total] | add)}'

# transform keys and values
jq '.data.orders | map(.total = (.total / 100))'
jq '.data.orders | sort_by(.createdAt) | reverse | .[0:5]'
jq '.data.orders | group_by(.status) | map({status: .[0].status, n: length})'

# exit status drives shell logic
if curl -s "$URL" | jq -e '.data.orders | length > 0' > /dev/null; then
  echo "orders present"
else
  echo "empty or unavailable"
fi

# export a table
jq -r '.data.orders[] | [.id, .status, (.total / 100)] | @csv' orders.json

# merge several files into one array
jq -s '.' page-1.json page-2.json page-3.json

# interpolate a shell value safely, without string concatenation
jq --arg id "$ORDER_ID" '.data.orders[] | select(.id == $id)' orders.json

JSONC and JSON5 config files

Several tools accept a relaxed dialect for configuration. They are not JSON, and a parser that complies with the specification will reject them - which is exactly what happens when you copy a JSONC snippet into an API request body.

// a JSONC file: valid for the editor, invalid for JSON.parse
{
  // the compiler options that matter for this package
  "compilerOptions": {
    "strict": true,
    "noEmit": true,
  },
}

// reading it in code needs a tolerant parser
import { parse } from "jsonc-parser";
const config = parse(await readFile("tsconfig.json", "utf8"));

# strip comments and trailing commas before sending it anywhere
npx strip-json-comments tsconfig.json | jq . > /tmp/clean.json

# JSON5 from the command line
npx json5 -s config.json5

# keep comments out of data files entirely
# config -> JSONC or YAML; payloads -> strict JSON, always

FAQ

Related

Syntax and types When to use something else

Last refreshed 2026-09-18.