A tool born from a simple need:sending prompts to an LLM, without sending your clients.

LLMs have changed the way we write reports, analyses and summaries. But there's a detail many teams accept by default: to get a useful answer, you end up pasting into a prompt data whose departure from the company is a problem — client contact details, accounting data, identifiers.

ONYRI Sanitize solves this by stepping in between you and the LLM. You paste your text, the tool detects the sensitive values, replaces them with neutral tokens, and gives you an anonymized version to paste into the LLM. When the response comes back, you paste it again into ONYRI Sanitize, which restores your original data — the trail is never lost, but it never left your machine.

The architecture is deliberately minimal. No backend has access to the data: everything lives in your browser, and the server only delivers the application's static files. That's also why it can be deployed in sensitive environments: there's no chain of trust to validate — just your browser and local JavaScript.

ONYRI Sanitize is a project by ONYRI Strategy, an independent consulting firm in data strategy and security.