Coming Soon
Coming Soon
A security service you can't inspect is a security service you have to trust. VettID's entire stack — the enclave code, the apps, the agent connector, even the cloud infrastructure — is public on GitHub. Read it, audit it, build it yourself, and verify that what we run is what we published.
AGPL-3.0 LicensedOpen source isn't a marketing choice here — it's load-bearing. VettID's trust model depends on cryptographic attestation: your app verifies the exact code running in the enclave before sharing anything. Attestation only means something when the code being attested is public. Each production vault release's measurements (PCR values) will be published in a signed manifest and our release log, and the build is reproducible: that is how you check that the enclave you're talking to was built from the source you can read.
This website, the account and admin sites, the member and admin APIs, every AWS CDK stack, and the design docs, including the vault messaging spec.
The vault: the Nitro Enclave supervisor and per-member vault where member secrets live, the host-side forwarder, the hybrid post-quantum crypto, and the reproducible image build.
The relay: mailboxes that hold end-to-end encrypted messages they cannot read, and the Go client.
The push gateway: wakes your phone through Google or Apple with a push that carries no content. It takes registrations only from vaults that prove by attestation that they run a published release, and it sees a phone’s push token and when a wake happens, not which member the phone belongs to or what the wake is for. In testing; not yet switched on for members.
The Android app — credential handling, vault pairing, and session crypto on your phone. In testing; not yet released.
The iOS app — the same member experience, built native for iPhone. Not yet released.
The desktop companion — time-limited, capability-scoped sessions where your phone stays in control. Not yet released.
The Agent Connector — the sidecar an AI agent will run to use its owner’s vault under LEASH: paired with the vault as an agent device, acting only under grants the owner signs and can revoke, never holding the vault’s keys. The code here predates the current vault and relay and is being rebuilt for them.
The integration layer that lets organizations run Service Vaults and connect with members without storing their data. In build; not yet a product.
Three approaches to patient-record lookup compared side by side: static credentials, IAM tokens, and VettID's organizational vault.
Lightweight Encrypted Agent Secret Handling — a proposed open standard for secure, auditable, owner-controlled secret access by AI agents. Framework-neutral: MCP is its first binding, and it complements MCP’s own authorization rather than replacing it.
Zero-Knowledge Trust: the next evolution of security for the agentic era — the thinking behind VettID's architecture.
Issues and pull requests welcome. Security findings go to security@vettid.org — not the public tracker.