Is my source code encrypted before it leaves my machine?

Yes. The Privatemode proxy runs locally and encrypts all code context before it leaves your machine. It also performs remote attestation to verify the integrity of the Privatemode service, ensuring end-to-end protection for every request.

Does Privatemode store or retain code from Tabby sessions?

No. Privatemode is designed to never learn from your data. All inference requests are processed confidentially using confidential computing, and no code, prompts, or responses are retained after processing.

Which Tabby features work with Privatemode?

Tabby's chat feature works fully with Privatemode via the openai/chat HTTP connector. Embedding models for code context retrieval can also be configured. Code completion support depends on whether Privatemode's completions endpoint supports the FIM (Fill-in-the-Middle) prompting style that Tabby uses for inline suggestions.

How does encrypted inference affect performance?

Privatemode uses confidential computing hardware to minimize overhead. The local proxy adds only milliseconds for encryption and attestation, so response times are comparable to standard cloud AI services while providing full data protection during processing.

Integrating Privatemode AI with Tabby

Tabby is a self-hosted AI coding assistant, but local models can't match cloud performance. Privatemode gives Tabby access to state-of-the-art cloud models while keeping your code fully encrypted.

Introduction

Tabby integration

Benefits

Why integrating Tabby

Introduction

Use cloud-based AI in Tabby without compromising data privacy

A screenshot of the tabby ML user interface

Self-hosted AI that needs better models

Tabby users self-host to keep code on-premises, but local models can't match the quality of cloud LLMs. Adding a cloud backend means sending your source code to third-party providers, undermining the privacy benefits of self-hosting.

Cloud models without the privacy trade-off

Privatemode bridges the gap. It provides an OpenAI-compatible endpoint backed by state-of-the-art models running inside confidential computing enclaves, so Tabby gets cloud-grade AI without exposing your code.

Encrypted even during inference

Your code context is encrypted before it leaves your machine, processed inside a hardware-enforced enclave, and never stored or used for training. This is enforced by confidential computing hardware, not just policy.

Benefits

Why use Privatemode AI with Tabby?

End-to-end encrypted code completion and chat

Integrating Privatemode into Tabby brings cloud-AI capabilities to your self-hosted coding assistant while keeping your code confidential through end-to-end encryption and confidential computing. Privatemode is designed to never learn from your data.

State-of-the-art model selection

With Privatemode, you can choose from state-of-the-art LLMs to power Tabby's chat and code completion, all running confidentially.

Native OpenAI-compatible integration

Tabby natively supports OpenAI-compatible HTTP model backends via config.toml. Point it at the Privatemode proxy and both chat and completion features work instantly.

How to get started

How to set up Privatemode with Tabby

Get your API key

If you don't have a Privatemode API key yet, you can generate one for free here.

docker run -p 8080:8080 ghcr.io/edgelesssys/privatemode/privatemode-proxy:latest --apiKey <your-api-key

Run the Privatemode proxy

The proxy verifies the integrity of the Privatemode service using confidential computing-based remote attestation. The proxy also encrypts all data before sending and decrypts data it receives.

brew install tabbyml/tabby/tabby

Install Tabby

To install Tabby on macOS using Homebrew, run the command below. This will download and set up the Tabby CLI so you can run and manage the Tabby server.

mkdir -p ~/.tabby 
cat > ~/.tabby/config.toml << 'EOF' 
[model.chat.http] 
kind = "openai/chat" 
model_name = "qwen3-coder-30b-a3b" 
api_endpoint = "http://localhost:8080/v1" 
api_key = "dummy" request_timeout_secs = 120 
max_input_tokens = 32768  

[model.embedding.http] 
kind = "openai/embedding" 
model_name = "text-embedding-3-small" 
api_endpoint = "http://localhost:8080/v1" 
api_key = "dummy"  

[server] 
host = "0.0.0.0" 
port = 8081 
completion_timeout = 30 
EOF

Create the following configuration file for Tabby

Create the configuration directory and then generate the config file that tells Tabby how to connect to Privatemode AI.

tabby serve --port 8081

Start Tabby

To start Tabby with the configuration you just created, run the following command launching the Tabby server.

Done

You've successfully set up Tubby with Privatemode.

FAQ

Frequently asked questions about using Privatemode with Tabby

Minimal changes. Tabby natively supports OpenAI-compatible HTTP model backends. You add a few lines to your config.toml pointing the chat model to the Privatemode proxy, and all requests are automatically encrypted and routed through confidential infrastructure.