We kept seeing the same pattern. Teams wanted to use GPT, Claude or Gemini on real customer data - support tickets, contracts, medical notes, claims. But every time someone pasted a name, ID's, SSN, email or account number into the prompt, the data left the building. Forever. Most “solutions” were either: “Just don’t paste sensitive data” (people still do), heavy self-hosted pipelines that take weeks to build, or tools that permanently redact everything and break the model’s ability to reason. We wanted something simpler. Change one field. The model never sees the real values. Your app still gets them back. That’s how OBVELO started. It’s a privacy gateway that sits between your application and any LLM provider. It replaces every personal value with a stable token before the request leaves, then restores the original values in the response. Same person keeps the same token, so the model can still follow the conversation. Nothing from the content is stored. We’re a small team from Kraków. We built this because we needed it ourselves - and because the existing options felt either too fragile or too heavy for most teams. Would love your feedback, especially if you’re dealing with this problem in healthcare, legal, finance or public sector.

OBVELO

Personal data stays with you - the question goes to the LLM

Personal data stays with you - the question goes to the LLM