Skip to main content

Overview

Enclava is an open-source AI platform with built-in confidential computing. You get OpenAI-compatible APIs, RAG, chatbots, and agents - but your data stays encrypted during processing through hardware-based trusted execution environments.

What Enclava does​

You can use it as a drop-in replacement for OpenAI's API, or build on top of its features:

RAG - Upload documents (PDF, DOCX, TXT, Markdown), and the platform chunks, embeds, and stores them in Qdrant. Your chatbots and agents can then pull relevant context when answering questions.

Chatbots - Create chat interfaces with custom system prompts and personalities. Each chatbot can connect to a RAG collection for context-aware responses.

Agents - AI that can use tools: search your documents, browse the web, run Python code, or call custom APIs via MCP servers.

Confidential inference - Requests route through TEE-based inference providers (Private Mode in EU, Redpill in US). Your prompts and responses stay encrypted even during GPU processing.

The interface​

The web dashboard handles most common tasks: uploading documents, creating chatbots, testing prompts, managing API keys. It's a Next.js app with light/dark mode.

For programmatic access, there are two APIs:

  • /api/v1/ - Public API with API key auth (OpenAI-compatible)
  • /api-internal/v1/ - Internal API with JWT auth (used by the frontend)

Getting started​

The fastest path:

git clone https://github.com/enclava-ai/enclava.git
cd enclava
cp .env.example .env
docker compose up

Then open http://localhost:3000 and create an API key.

If you already have an app using the OpenAI SDK, point it at Enclava's endpoint and it should work without code changes.

Next steps​