Is my workspace data encrypted before leaving my machine?

Yes. The Privatemode proxy runs locally and encrypts all data 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 data from my workspaces?

No. Privatemode never learns from your data. All requests are processed using confidential computing, and no documents, prompts, or responses are retained after processing.

Do all AnythingLLM features work with Privatemode?

Chat, document retrieval, and any other feature that uses the configured LLM provider route through Privatemode, and the embedder can be pointed at the same proxy. Agent skills depend on tool calling, which Privatemode's chat models support.

How does encrypted processing affect performance?

Confidential computing adds a small overhead. Response times are comparable to conventional AI services, and your data stays encrypted in memory during processing.

Integrating Privatemode AI with AnythingLLM

AnythingLLM turns folders of documents into workspaces you can question, with agents and a vector database behind them. Privatemode provides both the chat model and the embedding model behind those workspaces, so your documents stay encrypted in memory even while they are processed.

Introduction

Screenshot of AnythingLLM with Privatemode

The challenge

The AI sees your documents, not just your question

AnythingLLM answers a question by searching your documents and sending the matching passages to the model together with your question. In a workspace built from contracts, records, or personnel files, those passages are the sensitive part. With a conventional AI provider, they are processed in plaintext on the provider's servers, and so is every document while it is indexed.

The solution

Indexing and answers on a confidential model

AnythingLLM lets you choose the chat model and the embedding model separately, and both can point at the Privatemode proxy running on your machine or server. That covers the whole pipeline, indexing your documents as well as answering questions about them. Leaving the embedder on a conventional provider would send every document there at indexing time.

Benefits

Why use Privatemode AI with AnythingLLM?

Documents and chat both protected

See models

End-to-end confidentiality

Learn how

Free tier available

How to get started

How to set up Privatemode in AnythingLLM

Get your API key

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

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.

Set up AnythingLLM

Download AnythingLLM from the official website and open the application on your computer.

Configure AnythingLLM to use Privatemode

Open Settings, then AI Providers, then LLM, and choose OpenAI (Generic). Set the Base URL to http://localhost:8080/v1, any placeholder as the API Key, glm-latest as the Chat Model Name, and 1048576 as the Token context window. AnythingLLM does not fetch the model list for this provider, so the model name and context window are typed in by hand.

Point the embedder at the proxy too

The chat model and the embedder are configured separately. Open Settings, then AI Providers, then Embedder, select Generic OpenAI, and set the Base URL to http://localhost:8080/v1, any placeholder as the API Key, and qwen3-embedding-4b as the Embedding Model. Without this step every document would be sent through a conventional embedding API at ingestion time. Changing the embedder means re-embedding existing workspaces.

FAQ

Frequently asked questions about using Privatemode with AnythingLLM

No. AnythingLLM supports custom OpenAI-compatible endpoints natively via its OpenAI (Generic) provider. You set the Base URL to your Privatemode proxy endpoint and type in the model name and context window. Chat and document retrieval work as before.