Inside your environment
The Python kit signs a brief and encrypts it to a pinned recipient key using JWS and JWE. Transport stays with your existing authenticated queue or service.
Team-operated review workflows / Pilot
Give a reviewer agent the brief it needs, without handing it your entire client workspace.
Approved fields. One recipient. Short-lived access. Your keys and task payloads stay in your environment.
Synthetic agency brief / reviewer-01
No brief released.
No decision yet.
Simulation only: no encryption, signature, delivery, execution, or payment occurs here. The local kit performs encryption and signature verification. It does not run an agent.
The Python kit signs a brief and encrypts it to a pinned recipient key using JWS and JWE. Transport stays with your existing authenticated queue or service.
The receiver checks issuer, audience, allowed fields, scope, expiry, revocation and prior acceptance. Missing replay state blocks delivery.
A recipient-signed receipt records acceptance and a ciphertext fingerprint. It omits task content and is not proof of completed work or service quality.
Recruiting three pilot teams
$299 / one 30-day pilot
For agencies and internal teams running a repeated lead-agent to reviewer-agent workflow. Start with synthetic briefs, then decide whether the integration is suitable for your data.
Payment boundary: apply free. A manual invoice is issued only after written scope agreement and delivery of the agreed acceptance tests. No automatic renewal, production SLA, audit certification, or autonomous spending.
Pilot capacity is a recruitment target, not evidence of customers. This is not a hosted messaging product. If your workflow already has equivalent controls, keep them.
The kit uses the established JWCrypto library with fixed algorithms. This is application-level policy around standard cryptography, not a new encryption protocol.
Your operator provisions keys and authorization. An authorized recipient can copy plaintext. Revocation cannot undo a completed handoff; encryption does not solve prompt injection.
Keep existing transport authentication and authorization. There is no token passthrough, credential brokering, wallet custody, or claim of verified agent identity.
Inspect the implementation and threat model · Review the source and tests on GitHub · Need agent catalog distribution instead?