Beyond LLM chat: anonymize any text, documents and data

TokenVeil is often described as a privacy layer for AI chat, but the AI path is only one consumer of a general engine. The core is a reversible anonymize then deanonymize primitive that operates on text, whatever produces that text. The same primitive that protects a prompt to Claude also protects a document, a database export, a support ticket or a single field sent to any external service.

Any text, through the API

The public API takes arbitrary text, not only model prompts. You send text, you get a tokenized version plus an opaque session id, and you restore the real values later with that id. Nothing about it is specific to a language model.

• POST /api/v1/anonymize accepts a record, a field, a log line, a document’s extracted text or a full data export.
• POST /api/v1/deanonymize restores the real values from the session id, on your side.

The mapping between real values and tokens stays encrypted on your own instance and never leaves it. See the API reference for the full contract.

Documents and files

Beyond the API, TokenVeil ships a standalone file anonymization tool. You upload a file and get an anonymized, reversible result, without going through a chat.

• Text-based documents and structured files have their text extracted and anonymized.
• Scanned PDFs and images run through OCR first, so a scanned document is detected and anonymized through the same pipeline.
• The result is reversible: the original values are restored on your side when authorized.

A reusable layer across services

Because the engine works on text and exposes a versioned, UI-independent API, the same privacy capability is reusable well beyond AI tooling. Any internal service can tokenize sensitive data before it is processed anywhere external, and restore it afterwards:

• a developer tool sending code and context to a hosted model,
• a data pipeline sending records to an external processor,
• a document workflow sharing files with a third party,
• a support or analytics tool that must not expose customer data downstream.

The developer tools case is usually the sharpest place to start, but the /anonymize then external processing then /deanonymize pattern is the same everywhere. It is a stable, versioned integration surface (/api/v1): v1 keeps working as the product evolves.