RedPill is a confidential AI inference platform that enables users to verify the integrity and privacy of every AI response. Each inference request is run inside a Trusted Execution Environment (TEE), and the platform provides attestation reports and signed receipts that bind the user's request and the resulting response to the attested workload. This design allows users to independently verify evidence of confidential processing, ensuring that sensitive data is handled securely and transparently.
The platform features an ACI gateway that emits evidence artifacts such as nonce-bound attestation reports, workload keysets, and user-visible receipt IDs. Channel binding is enforced through downstream TLS SPKI or ACI end-to-end encryption public keys, which are published in the attested keyset, allowing clients to reject mismatched connections. RedPill also provides signed receipts containing details like request hashes, selected routes, upstream verification results, response wire hashes, and attested audit sessions. Provider adapters verify upstream targets including Phala, Tinfoil, NEAR AI, Chutes, and ACI/DCAP, enforcing secure bindings before forwarding requests.
Developers can integrate RedPill with their existing OpenAI SDKs by changing the base URL and selecting a confidential model, while maintaining the ability to verify the gateway attestation, response receipt, and attested session. The API supports a range of functionalities including chat, embeddings, audio, images, files, batches, and models. Additional features include E2EE routing, virtual keys with configurable budgets and rate limits, and an open-source gateway and verifier. A verifier CLI is available for independent proof checks.
RedPill is positioned for developers, legal professionals, and individuals who need to build privacy-first applications, protect regulated workflows, or provide confidential chat as a proof-of-product. It is especially relevant for use cases in healthcare, legal, or fintech sectors where compliance and confidentiality are critical. The platform offers access to over 200 models, including private and public large language models deployed in GPU TEEs, with transparent per-model pricing for input, output, and cache reads. These models support advanced features such as large context windows, multimodal input, configurable reasoning modes, and strong multilingual capabilities.
In the AI & ML space, RedPill takes a focused approach. It focuses on enabling verifiable, confidential AI inference with attestation and auditability for sensitive data. RedPill is a B2B product aimed at developers and enterprises needing secure AI inference. The product ships for API.
RedPill builds and maintains RedPill, and the product first shipped in 2026. The project is developed in the open on GitHub with 18 stars and 168 commits in the last 90 days. Across PulseGate's embedding index, RedPill has few near neighbours, marking it as relatively distinct. Among its 5 catalogued features are confidential inference, attestation reports, and signed receipts. It exposes integrations via a public API.
Latest indexed changes and source events
Other apps tracked under the same category.
