Signed attestation and action log.
Enterprises need to know - and prove - what an AI agent did in their environment. A screenshot or a log line isn't evidence; it's a claim.
Evidence, not log lines
Proof produces a signed attestation for agent actions: an Ed25519-signed, append-only log that can be independently verified without trusting Surfil's servers at the moment of the claim.
Proof, not promises
Proof produces one metric type on the shared spine: signed attestations. Every paid output is signed (Ed25519) and verifiable offline with no account. Zero-trace: your source never leaves the device.
✓ VALID (offline · epoch 7)
Proof, as you'd actually see it
How Proof does it
Where Proof earns its place
Regulated environments
Prove what agents did - and didn't do - without asking anyone to trust a vendor dashboard.
Incident review
Reconstruct agent activity from signed entries instead of mutable logs.
Third-party assurance
Hand an auditor the chain and the public key; they verify it themselves.
Questions developers ask first
What makes this evidence rather than logs?
Signatures and append-only chaining. A log line asserts; an attestation proves.
Do verifiers need a Surfil account?
No. Verification is offline against published epoch public keys.
What if a signing key is compromised?
Keys rotate by epoch; historical signatures stay valid and rotation is dual-controlled.
One spine - products compound
Add Proof to your agents.
Core installs with Starter; Proof plugs into the same interception point - no second layer, no new setup.