Privacy-First AI Infrastructure Running Entirely in TEE
RedPill is an open-source AI platform that provides hardware-enforced privacy for all AI workloads. Unlike traditional AI gateways, RedPill's entire infrastructure runs inside Trusted Execution Environments (TEE), ensuring cryptographic guarantees that your data remains private.
|
Entire gateway runs in TEE (Intel TDX) |
Native TEE models with GPU secure enclaves |
Cryptographic attestation for all executions |
|
OpenAI, Anthropic, Google, Meta, DeepSeek |
Drop-in replacement for existing code |
Battle-tested infrastructure |
The core AI gateway that routes requests to 250+ LLMs while running entirely in TEE.
What it does:
- Routes AI requests through hardware-protected TEE
- Supports 250+ models from all major providers
- Provides cryptographic attestation
- OpenAI-compatible API
Tech: TypeScript, Hono, Cloudflare Workers, Node.js
npm install && npm run dev:nodeModern, privacy-focused chat interface with cryptographic verification.
What it does:
- Beautiful chat UI with Assistant UI components
- Real-time TEE attestation verification
- Message signature verification
- 250+ model support
Tech: Next.js 15, React 19, Vercel AI SDK, Tailwind CSS 4
npm install && npm run devPython tools for validating TEE attestations and response signatures.
What it does:
- Verify GPU TEE attestation (NVIDIA H100/H200)
- Validate Intel TDX quotes
- Check ECDSA signatures
- Sigstore provenance verification
Tech: Python 3.10+
pip install -r requirements.txt
python3 attestation_verifier.pyfrom openai import OpenAI
client = OpenAI(
api_key="YOUR_API_KEY",
base_url="https://api.redpill.ai/v1"
)
response = client.chat.completions.create(
model="openai/gpt-5",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)# Generate fresh nonce
NONCE=$(openssl rand -hex 32)
# Get cryptographic attestation
curl "https://api.redpill.ai/v1/attestation/report?nonce=$NONCE" \
-H "Authorization: Bearer YOUR_API_KEY"Returns cryptographic proof:
- ✅ GPU TEE measurements (genuine NVIDIA H100 TEE)
- ✅ Model hash (exact model version)
- ✅ Code hash (inference code integrity)
- ✅ Hardware signature (signed by TEE)
Layer 1: TEE-Protected Gateway (All 250+ Models)
Your Request → Intel TDX TEE → Any LLM → Intel TDX TEE → Your Response
- Request processing in CPU TEE
- Response handling in CPU TEE
- Works with all providers
- No additional cost
Layer 2: TEE-Protected Inference (Phala Models)
Your Request → Intel TDX TEE → NVIDIA GPU TEE → Intel TDX TEE → Your Response
- Model weights in GPU secure enclaves
- Inference computation in GPU TEE
- Complete end-to-end protection
- Cryptographic attestation
- NVIDIA GPU vendor (H100/H200 TEE correctness)
- Intel CPU vendor (TDX implementation)
- Phala Network (model deployment integrity)
- Open source code (auditable on GitHub)
- RedPill operators
- Cloud provider (AWS, GCP, Azure)
- System administrators
- Other users on same hardware
|
🏥 Healthcare
💰 Financial Services
⚖️ Legal
|
🏢 Enterprise AI
🏛️ Government
🔬 Research
|
- 📖 Docs: docs.redpill.ai
- 🔐 Attestation Guide: Verify TEE Protection
- 🔑 Verification Guide: Signature Verification
- 🏗️ Architecture: TEE-Protected Gateway
We welcome contributions to all RedPill projects!
- Fork the repository
- Create a feature branch
- Make your changes
- Run tests and linters
- Submit a Pull Request
See individual repository CONTRIBUTING.md files for detailed guidelines.
- 🌐 Website: redpill.ai
- 📖 Documentation: docs.redpill.ai
- 💬 Discord: discord.gg/redpill
- 📧 Support: [email protected]
- 🐦 Twitter: @redpill_ai
- Phala Network - TEE infrastructure and confidential AI models
- NVIDIA - H100/H200 GPU TEE technology
- Intel TDX - CPU TEE technology
Built with 💜 by the RedPill team
Making AI privacy-first, one request at a time.
