Map any JSON to any JSON — in minutes, not sprints.

Paste one sample record from each system. Schemble proposes the field mapping with Claude, you review it and preview it on your samples, then export the mapping plus dependency-free TypeScript and Python transformation code.

Private beta — access is password protected.

Legacy CRM
{
  "CustomerID": "10042",
  "FirstName": "Ayşe",
  "LastName": "Demir",
  "SignupDate": "14/03/2021",
  "Status": "A",
  "IsVip": "yes"
}
New platform API
{
  "id": 10042,
  "fullName": "Ayşe Demir",
  "createdAt": "2021-03-14T00:00:00Z",
  "status": "active",
  "vip": true
}

How it works

  1. Paste samples. One record (or a few) from the source system and from the target system.
  2. Get a proposed mapping. Claude matches fields and picks transformations, with a confidence score and a reason for every rule.
  3. Review and preview. Edit any rule, then run it on your samples and see exactly which target fields match, are missing or need attention.
  4. Export. Download the mapping file and generated TypeScript or Python code you can drop into your integration.

Built for integration work you can trust

Your data stays yours

Only field names, types and detected formats (such as date:DD/MM/YYYY) are sent to the AI — never your sample values. Nothing is stored on our servers.

Deterministic code

The AI proposes; it never writes your code. Transformation code is generated from the reviewed mapping by templates, with no runtime dependencies.

A human in the loop

Low-confidence rules are flagged for review, unmapped values are highlighted, and the preview checks the output against your target sample.

Transformations

Copy, type conversion, date and timestamp formats (including ISO 8601 and Unix time), concatenation, splitting, value maps for codes, constants, defaults, and per-item mapping of nested arrays.