Keep a user/order join in a shareable JSON example.

1. Choose approved examples

Open the workbench and select “Load users + orders example.” All records are synthetic. Your own inputs stay in page memory; the downloaded HTML works offline. There is no automatic sensitive-field discovery.

2. Select paths deliberately

The first input contains two users; the second contains orders referring to those IDs. Apply these rules:

[
  {"path":"/users/*/id","action":"token","namespace":"USER"},
  {"path":"/users/*/email","action":"replace","value":"example@example.test"},
  {"path":"/orders/*/userId","action":"token","namespace":"USER"}
]

One asterisk matches one level. The two ID rules share USER, so equal source IDs share tokens across both documents. Emails use a fixed example value. Order IDs and totals remain unchanged.

3. Run and inspect both documents

Expected: six selected fields across two documents; no unmatched rules in the whole batch. User alpha becomes USER_001 and beta becomes USER_002. The first order refers to beta, so its userId is USER_002. Switch the current-document selector to inspect the second result.

Inspect the remaining totals, IDs and any free text. The manifest reports paths, actions and original types without a reversible token map or source values. Field names and paths can still identify someone.

4. Approve and export

Confirm the review only after inspecting every document. Download each current JSON and the batch manifest. Input or rule edits immediately clear outputs and approval. Tokens reset on another run; include related files together and keep order stable for reproducible names.

Try the paired example Get the offline sample

5. Repeat from the command line

Extract the free sample and install/use Node.js 22+ if needed. No package installation is required. Run from the extracted folder:

node cli.mjs recipes/01-users-orders/rules.json reviewed-output recipes/01-users-orders/input-1.json recipes/01-users-orders/input-2.json

The output folder must be new. Compare the two generated documents to the recipe's expected files. The runner validates the whole batch before creating output. A fresh-folder check is not proof of support on every machine.

Choose a suitable alternative

Pino redaction is a free alternative for application logging. Presidio structured workflows address detection and anonymization workflows. Mockaroo generates new synthetic data and has a free tier. Tonic offers broader synthetic-data products.

This workbench is a narrow local review interface for approved JSON examples. It does not detect every sensitive field, encrypt data or certify a regulatory outcome. Current competitor pages were checked on 7 October 2026; pricing can change.

What the experimental kit adds

The fully usable free tool and paired exercise remain free. The prepared $9 recipe kit adds eleven more synthetic workflows, expected outputs and review notes. Checkout is closed while the owner handles payment setup. No buyers, testimonials or revenue are claimed.