Authorize access.
Your client asks account services to check access and reserve credits. This request contains the model and usage limits, without your prompt.
Privacy, built in.
Use the models you love from your terminal, an SDK, or your browser. AnonRouter routes your prompts through protected hardware, without logging them or using them for training.
01 / Protected processing
AnonRouter handles your requests inside an Intel TDX enclave. It runs on the server, with memory protected from the host operating system and hypervisor.
The host operating system manages the server. AnonRouter runs in an enclave on that same machine, with memory that Intel TDX isolates from the host operating system.
Your encrypted connection opens inside the enclave. AnonRouter handles your prompts and responses there, without writing them to logs.
02 / Your identity stays separate
Create an anonymous account and connect from your terminal, SDK, or browser. Account services authorize access and handle billing without receiving your prompts.
Your client asks account services to check access and reserve credits. This request contains the model and usage limits, without your prompt.
Your client receives a short-lived ticket authorizing the request. It can be redeemed once, replacing account credentials on the content request.
Your client sends the ticket and prompt to the protected endpoint without account credentials. Account services receive usage metadata for billing, not your conversation.
Using a terminal, coding agent, or SDK? With a standard OpenAI-compatible client, you send your API key and prompt over HTTPS to the enclave. Authentication happens there; account services receive only authorization and billing metadata. You don’t need to request a ticket yourself.
Create a private account03 / Choose your protection
Our no-logs promise applies across AnonRouter. Route labels show the additional protection from the provider, hardware, and encryption.
The provider can see your prompt while serving it, but your identity stays hidden.
Prompt and response content is not retained after the request completes.
The model runs inside an attestable enclave, isolated from the infrastructure around it.
Your client encrypts the prompt; only a verified enclave can decrypt it.
Anonymous routes hide your AnonRouter identity from the provider; the provider’s own content retention policy applies. Private routes add a commitment not to retain prompts or responses after the request.
The model runs inside an attestable enclave, with memory isolated from the host software. E2EE adds encryption from your client to the verified enclave, so only that enclave can open your prompt.
04 / Verify it yourself
Explore the public source, inspect your route’s attestation, and check the evidence behind its protection.
The confidential content plane is public. Read the code that handles your prompts inside the enclave, explore its tests, and inspect how the workload is built.
Explore the sourceRun verification from your terminal or application with the Confidential SDK. Check AnonRouter’s routing enclave and your chosen provider route separately, then make your request only when the results meet the protection you require.
Explore the Confidential SDKInspect the verification results and compare the measured workload with the expected release. Use the provider’s tools to check its hardware evidence independently. In web Chat, the attestation panel also lets you download the raw evidence.
Read the verification details05 / What remains afterward
Conversation content, usage records, and local history are different things. Here’s what happens to each.
AnonRouter doesn’t store your prompts or responses or use them for training. The model provider’s retention depends on your chosen route; Anonymous routes follow the provider’s own policy.
We retain limited metadata such as the model, token counts, cost, and request status to bill for usage and operate the service. Your prompt and response are excluded.
Terminal tools, coding agents, and SDK applications have their own history and logging settings. Our web Chat and Studio can save history in your browser; clear it, disable it, or use Ghost Mode to keep conversations out of local history storage.
Use the models you love. Keep your conversations yours.