Introduction
Transmutify is a headless API that turns messy spreadsheets and unstructured text into clean, schema-shaped records — with hosted review before anything is processed.
You define a target schema once: the fields your system expects, their types, and which ones are required. From then on, every client spreadsheet, export, or raw text snippet can be transformed into exactly that shape with a few API calls — no embedded widget, no manual column mapping UI for your team to build.
See it in action first: the Playground in the dashboard runs the full flow — upload a file, watch the AI prepare the mapping, and open the hosted review — without writing a line of code. After the review redirects back, the callback page shows compact, copyable cURL examples for the run-specific API calls that created the review. And the ✨ AI Instructions tab hands your coding agent (Claude Code, Cursor, …) a ready-made prompt to plan and build the entire integration with you.
How it works
- Create a transformation run against one of your schemas — either for a file upload or for raw text content.
- Upload the source file to a pre-signed URL (file flow only). Transmutify inspects the file and detects headers automatically.
- Review the mapping. An AI provider of your choice proposes how source columns map to your target fields, with generated examples shown inside the hosted Review UI.
- Approve the review. Transmutify processes every row and produces a JSON or CSV artifact you can download.
The raw content flow still uses hosted Review: submit up to 50,000 characters of unstructured text, then reviewers validate and edit the generated rows before approval. See Raw content flow.
Choose your flow
| Flow | Input | Provider | AI data exposure | Review starts at | Best for |
|---|---|---|---|---|---|
| Structured file View guide | CSV, TSV, TXT, or XLSX | Optional | Sanitized headers and sample rows only | Mapping, then Validate Data | Spreadsheets and exports with many rows |
| Image extraction View guide | JPG, PNG, or WebP | Required | Uploaded image is sent to the provider | Validate Data | Screenshots, scans, and photographed records |
| Raw content View guide | Short unstructured text | Required | Submitted content is sent to the provider | Validate Data | Email signatures, notes, and pasted snippets |
| Dry run View guide | Any create request shape | Request shape only | No real data or provider call | Same lifecycle with synthetic data | Staging, automated tests, and integration wiring |
Key concepts at a glance
- Schemas describe your desired output: field names, descriptions, data types, and required flags.
- Providers are the AI configurations (Anthropic, OpenAI, xAI, Gemini, OpenRouter) your organization registered for preparing mappings and extracting reviewable output.
- Transformation runs track a single transformation through its lifecycle — from
awaiting_uploadtocompleted— across a review and a final stage.
Read more in Core concepts and Review flow.
Where to go next
- Quickstart — make your first transformation in a few minutes.
- Authentication — create an API key and authenticate requests.
- API Reference — every endpoint, request field, and response shape.