01
Your agent signs the canonical payload.
On your machine.
THE CRYPTOGRAPHIC TRUST LAYER FOR AI AGENTS
Your agent signs the bytes that matter on your own machine, with a key we never hold. We verify the signature against a public key you uploaded. That’s the entire trust story.
THE CLAIM
Your agent constructs a canonical payload — the few bytes
that matter for trust (the request ID, timestamp, recipient,
permissions, and a hash of the content). It signs those bytes
with an SSH key you generated on your own machine, using the
standard ssh-keygen -Y sign flow over SSHSIG. We
verify the signature against a public key you uploaded to us
— your private key never leaves your machine.
Ed25519 by default; ECDSA-P256/384/521 and RSA-3072+ supported.
Wire format is open; conformance vectors are publishable.
HOW IT WORKS
01
Your agent signs the canonical payload.
On your machine.
02
We verify the signature against your public key.
We never see the content.
03
Your verifier gets a clear verdict.
Verified or not. With a reason.
The “canonical payload” is the set of fields we sign for trust — request ID, timestamp, recipient, permissions, and a hash of the content. Not the content itself.
FROM ONE TERMINAL
$ apsign sign --key ~/.ssh/id_ed25519 < payload.json \ | apsign verify --token av_live_… \ --payload payload.json --signature - verified: true key: SHA256:abc123…wxyz
preview — apsign CLI in development; v0.1
coming month 4. Today, the same signature flow works with
one ssh-keygen -Y sign invocation. Spec available
on request.
DESIGN PARTNERS
DESIGN PARTNERS · COHORT OF FOUR
Names go up when our partners are ready to be named. Until then, the work speaks.
Apply to the cohort →If you’re building an agent that needs to prove its messages are real, we want to hear from you.
ACCESS
We’re working with a small number of teams shipping agent products in production. Pricing is bespoke during the design-partner phase.
Email founder@agentsproof.ai →