Introduction

Privatemode is an inference API for open-weights AI models with end-to-end encryption. Prompts and responses are processed only inside hardware-isolated confidential computing environments. Neither Edgeless Systems as the operator nor the infrastructure provider can access them, and your client verifies this through remote attestation before it sends any data.

The API is compatible with the OpenAI and Anthropic APIs, so existing applications, SDKs, and tools work with minimal changes. Privatemode is operated by Edgeless Systems, a company based in Germany, and runs on confidential-computing infrastructure in the EU.

How you connect

  • JavaScript and TypeScript: the Privatemode SDK handles attestation and encryption inside your application.
  • Other languages and tools: the Privatemode proxy runs next to your application and exposes a local OpenAI-compatible and Anthropic-compatible endpoint. Coding assistants, agent frameworks, and the official OpenAI and Anthropic SDKs connect to it unchanged.
  • Easy evaluation: the proxyless API lets you try the API without additional client software.
  • In the browser: the web app lets you try the models without any setup. It's a convenient way to test models and prompts, not a substitute for the API.

Get started

Find your way around

The documentation has four sections, reachable from the top navigation:

SectionWhat you find there
Documentation (this section)Concepts and configuration: models, pricing and rate limits, API features such as prompt caching and structured outputs, the web app, the portal, security, and architecture.
CookbookStep-by-step recipes: coding assistants, tool calling, vision input, speech-to-text, and verifying the service from source.
ReferenceThe API reference for every endpoint, and the JavaScript SDK reference.
ChangelogThe release notes for every version of the service.

Evaluating Privatemode

If you assess Privatemode for your organization, start here: