GoMask API · for coding agents

Test data with the story already in it. Ask your agent for it.

One sentence in Claude Code, Cursor or Codex. Related tables, real patterns, in your repo.

llms.txt

10 to 1,000,000 rows · SQL, CSV, Parquet, JSON · from $0.25 per 1K rows

› you, in Claude Code

Use GoMask to seed staging with customers and their orders1. Orders over $500 are 14× likelier to be fraud2.

✓ signed in (you approved TJWR-FDXP)✓ seed/customers.sql 2,000 rows✓ seed/orders.sql 7,412 rows
1 · every key resolvesorders.customer_id → customers.id
2 · fraud 7.0% over $5000.5% under $50

Faker fills columns. GoMask fills them with a pattern.

SELECT amount_band, avg(is_fraud) FROM orders GROUP BY 1;
Faker script

Nothing for your fraud rule to catch.

GoMask

Big orders are where the fraud is.

An example of the same prompt, both ways.

Things your agent can ask for.

  • customers with their ordersTables that join
  • 3% of payments failRare events, at your rate
  • spend rises with seatsColumns that move together
  • 5% missing emails, some duplicatesMess to test validation
  • use real ICD-10 codesReal values, looked up and cited

Three calls. Or one sentence.

  1. 1
    Sign in once

    Your agent shows a code; you approve it in the browser.

  2. 2
    Preview free, then generate

    Check the schema, then run it at any size.

  3. 3
    Download

    Signed links, no auth header, valid for a day.

Full reference in llms.txt →
# paste into Claude Code, Cursor or Codex
Use GoMask to seed staging with customers
and their orders. 3% of payments fail.
Docs: datafactory.gomask.ai/llms.txt
 
# it reads the docs, asks you to approve,
# then makes the calls in the other tabs.

Pay per row. Previews free.

10,000 rows
$12.99
50,000 rows
$24.99
200,000 rows
$59.99
1,000,000 rows
$249.99

Pay once, rows never expire. See pricing

You stay in charge.

  • Only you can approve a sign-in
  • See and revoke every agent key
  • Runs stop at your balance, never a surprise
  • Every row synthetic, safe to commit

Questions

When should I just use Faker?

When the values do not matter. When your code branches on the data (fraud rules, churn logic, dashboards, search), the pattern has to exist, and independent random columns will not have it.

Is there an SDK?

There is no need for one: it is a handful of JSON endpoints, and llms.txt is written so your agent can follow it on its own.

How long does a run take?

The schema drafts in seconds. Larger runs keep going in the background while your agent polls, so nothing holds a connection open.

Does it use my data?

No. You send a sentence, never a database. Every row is synthetic, so it holds no real people and is safe to commit, share in a demo or load into staging.

What should your test data show?

Edit the sentence, paste it into your agent. Or preview 20 rows free in the browser.