TrustedRouter is an AI gateway for verifiable confidential computing. Its core purpose is to let users access hundreds of models through a single OpenAI-compatible API while controlling providers and privacy standards during routing.
Main capabilities:
- Single API access to 659 models, all routes are labeled, prices are shown per million tokens, starting price is the lowest-price route, and the degree of privacy varies by route.
- Verifiable confidential computing: the gateway runs in a trusted execution environment (TEE), publishes build attestation, links the running image to public source code; it provides signed receipts and supports verifying evidence on the trust page.
- Open source: the gateway, SDK, and deployment configuration are public, and attestation links the source code to the code handling requests, so users can read the source, inspect attestation, match digests and commits, and verify its fail-closed behavior.
- Privacy protection: prompts and output content deliberately do not enter the control plane and metadata systems; optional content export requires explicit target-level configuration.
- Routing control: users can control which providers handle requests and the privacy standards that must be met; if no route meets the minimum privacy requirements, the request stops before inference and will not be silently downgraded.
- Multi-provider and health monitoring: across three clouds, continuously measures provider and model health, and publicly discloses latency and availability history.
- Budget control: strict budgets count still-running requests toward each key's daily, weekly, or monthly limits.
- BYOK: supports bringing your own key, preserving provider commitments or enterprise rate limits while using TrustedRouter's routing and trust boundary.
How to use: Keep your existing OpenAI SDK, change only the base_url, and continue using the same chat or Responses calls; the migration guide includes tested examples. Add privacy and routing constraints only when needed. You can start without a credit card, with 1 million free tokens included. Users need to choose a login method to enter the workspace, and new users can create an account.
Use cases:
- Developers who need to route across multiple models and providers while requiring verifiable privacy and confidential computing.
- Compliance or security-sensitive applications that need audit and attestation evidence.
- Existing OpenAI applications that want to migrate to more private routes with minimal code changes.
- Case: RRH uses one OpenAI-compatible API for extraction, retrieval preparation, and multi-stage legal reasoning, processing 170,974 legal documents, billions of routed tokens, reaching production within three weeks, with every requestthrough TrustedRouter's attested gateway.
Limitations and notes:
- The degree of privacy varies by route, and not all routes provide the same privacy standard.
- The fail-closed mechanism means that if no route meets the minimum privacy requirements, the request stops directly, which may affect availability.
- Prices, specific model list details, the specific conditions of free tokens, etc. are subject to the official page; the materials do not provide a complete pricing table or a list of all models.
- More details about accounts, terms of service, and privacy policy are not yet provided.
