Analyze sensitive evidenceinside an attested AI enclave.

Mercy turns messy case files into structured outputs while the workload runs inside a TEE, proves itself with remote attestation, and only receives case and model keys after secure release policy passes.

Evidence files

Entities

Review flags

Timeline events

TEE execution

Policy key release

Mercy Graph

Live case graph: Project Nightingale

TEE execution boundary

Attestation before trust

Keys released by policy

The holy grail

The hard part is not the model. It is proving who gets the keys.

Mercy makes confidential AI legible: the workload boots inside a trusted execution environment, remote attestation proves exactly what launched, and secure key release unlocks sensitive material only for approved measurements.

Sensitive prompts, retrieved evidence, and model context stay inside the enclave while the workload is running.

Remote attestation proves the exact runtime before the system is trusted with regulated data.

Case keys and model secrets are released only when enclave measurements satisfy policy.

Every run leaves behind an audit trail that investigators, security teams, and customers can review.

The trust chain, made visible.

Boot inside a TEE

Each AI workload starts in hardware-isolated memory so prompts, retrieved evidence, and model context stay protected while in use.

Verify remote attestation

Mercy checks measurements, runtime identity, and policy before treating the workload as trusted.

Release keys only on proof

Case secrets, retrieval keys, and model credentials are unsealed only for approved enclave measurements.

Ship outputs with an audit trail

Analysts keep the case workflow fast while security teams get the release and attestation record behind every run.

Built for internal demos

Open the prototype now.

Everything is available without a request-access gate so you can walk teammates from the trust boundary to the case output in one sitting.

TEE execution

Remote attestation

Policy key release