Private AI Gateway · verified providers · signed receipts

BuildPrivate AI.

Add end-to-end private AI with one OpenAI-compatible gateway. Route across verified TEE providers, optimize price and speed, and keep a signed receipt for sensitive responses.

Get API Key Documentation Gateway GitHub

Private AI Gateway Features

Smart Routing

Choose availability, price, or end-to-end encrypted routing. Private mode never downgrades a protected backend to a standard fallback.

Virtual Keys

Create sk-rp-* keys with per-key budgets, RPM and TPM limits, and routing strategy controls for each app or customer.

Verification & Receipts

Call /v1/attestation/report with a nonce, then fetch the signed receipt through /v1/aci/receipts/{id} when you need proof.

End-to-End Encrypted AI

E2EE v2 uses AES-256-GCM, secp256k1 key exchange, nonce binding, and timestamp replay protection.

Attested Audit Sessions

Receipts reference read-only audit sessions so relying parties can check target, verifier, evidence digest, and channel binding.

Full API Surface

Chat, completions, embeddings, and model discovery work the same way you already call them; RedPill middleware extends the broader API surface.

Browse Private Models

Up and running in minutes

YOUR SDK.
PRIVATE AI.

Keep the OpenAI client you already use. Change the base URL, choose a private model, and add verification when your workflow needs attestation reports, response receipts, and audit sessions.

View Full Docs Private AI Gateway RedPill GitHub

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://tee.redpill.ai/v1",
  apiKey: process.env.REDPILL_API_KEY,
});

const response = await client.chat.completions.create({
  model: "openai/gpt-oss-120b",
  messages: [{ role: "user", content: "Say hello." }],
});

Open Source Where It Matters

Audit the gateway, run the verifier, inspect the chat UX, and track signed receipts as a developer-preview proof layer.

Private AI Model Catalog

The most private model options in one place, with unified pricing, provider coverage, and live performance through one gateway.

Z.ai: GLM 5.3New

GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks. It supports text input and output with a 1M-token context window. Served as a text-only TEE deployment via Phala.

Aug 31, 2026
1M context
$1.40/M input
$4.40/M output
$0.26/M cache read

Z.ai: GLM 5.3 FlashNewBeta

GLM-5.3-Flash is Z.ai's natively multimodal 320B MoE model with 18B active parameters, designed for efficient coding, long-horizon agent tasks, visual understanding, and long-context inference. Served as a TEE deployment via Phala.

Aug 27, 2026
1M context
$0.15/M input
$0.50/M output
$0.03/M cache read

Qwen: Qwen3.8 27BNewBeta

Qwen3.8 27B is an open-weight dense vision-language model from Qwen. It is suited for coding, professional workflows, research, multimodal interaction, and long-running agent tasks, with flexible thinking that can be enabled or disabled. Served on Phala in a TDX-attested enclave.

Aug 24, 2026
262K context
$0.30/M input
$3.00/M output
$0.05/M cache read

Meta: Muse Glimmer 30BNewBeta

Muse Glimmer 30B is a dense, open-weight multimodal model from Meta Superintelligence Labs, distilled from Muse Spark and optimized for autonomous agents on consumer hardware. It is suited for long-horizon agentic and coding workflows, with multi-step reasoning, reliable tool use, failure recovery, image understanding, and multilingual support across more than 100 languages.

Aug 13, 2026
131K context
$0.30/M input
$1.10/M output
$0.04/M cache read

DeepSeek: DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows.

Aug 4, 2026
1M context
$0.44/M input
$1.32/M output
$0.028/M cache read

Phala: Gemma-4 26B-A4B Uncensored (Heretic)

Uncensored "Heretic" variant of google/gemma-4-26B-A4B-it created using Heretic v1.2.0 with the Arbitrary-Rank Ablation (ARA) method and row-norm preservation. Refusals drop from 100/100 to 11/100 with KL divergence 0.0499 vs the base model. The base Gemma 4 26B A4B is a Mixture-of-Experts model with 25.2B total / 3.8B active parameters (8 active / 128 total experts), 30-layer transformer with hybrid local sliding (1024) + global attention, supporting a 256K context window. Natively multimodal (text + images, variable aspect ratios). Strong on coding, reasoning, function calling, with native system prompt support across 35+ languages. Served on Phala in TDX-attested H200 enclave with end-to-end ECDSA response signing; vLLM-compatible FP8-Static quantization by cloud19 (router excluded from quantization).

May 23, 2026
66K context
$0.15/M input
$0.70/M output

Browse Private Models

Start Building Private AI.

API Documentation

Comprehensive guides, API references, and tutorials to help you integrate RedPill into your applications. Add encrypted paths, private model routing, and verification without changing your SDK.

View Documentation Gateway GitHub

Developer Community

Join our Discord community to connect with other developers, get help with integration questions, and share what you're building with RedPill.

Join Discord

Get an API key in 2 minutes

Start with the OpenAI SDK, choose a private model, and add verifier checks when you need audit-grade proof.

Get API Key

Private Chat

Private model

Hi. Ask me anything and I'll open the confidential chat with openai/gpt-oss-120b selected.

Zero data retentionVerifier ready