Cookbook

Recipes for building with Privatemode

Step-by-step guides that show how to employ Privatemode in real-world scenarios, including confidential coding assistants, audio transcription, and verifying the services from source code using remote attestation.

Coding assistants

Coding assistants

Use Privatemode for AI-assisted coding on sensitive codebases: supported models, IDEs, and extensions including Copilot, Claude Code, and Cline.

Read recipe

Claude Code

Set up Anthropic's Claude Code with Privatemode: configure the proxy, set environment variables, and start confidential AI-assisted coding.

Read recipe

OpenCode

Configure OpenCode to use Privatemode as a model provider: set up the proxy, add the custom provider configuration, and start coding confidentially.

Read recipe

Visual Studio Code

Set up Privatemode in VS Code with GitHub Copilot, Continue, or Cline: private AI chat, editing, and tab completions for sensitive code.

Read recipe

JetBrains IDEs

Set up Privatemode in JetBrains IDEs: configure the AI Assistant plugin with IntelliJ, PyCharm, or CLion for private AI coding.

Read recipe

Audio

Speech-to-text

Transcribe audio with Privatemode's confidential speech-to-text: choose between Whisper and Voxtral, set the language, get timestamps, and handle long files.

Read recipe

Vision

Vision input

Using vision (image) input on Privatemode with DeepSeek-OCR-2, and Kimi K2.6.

Read recipe

API

Tool calling

Implement encrypted tool calling (function calling) with Privatemode: define functions, handle responses, and keep data confidential.

Read recipe

Verification

Verification from source code

Verify Privatemode's trust chain from source code: reproducibly build container images, generate the Contrast manifest, and compare hashes.

Read recipe

Verification of model integrity

Verify Privatemode model integrity: reproduce dm-verity protected model disks and compare root hashes against the deployed configuration.

Read recipe