CNothing AI Registration Hub Standard 1.0
Status: Public Architecture Standard
Published: 2026-04-06
Canonical Path: /standards/registration-hub
This standard defines how CNothing is used as the authentication, credential-protection, and recovery center of an AI-operated website registration system in which agents can orchestrate signup flows without learning user secrets.
1. Scope and Purpose
Defines the architecture problem this standard solves for AI-operated signup systems.
This standard specifies how CNothing participates in an AI-driven website registration system as the central trust service for client identity, protected onboarding data, issued credentials, and recovery-safe state transitions.
The purpose of the architecture is to let an AI agent coordinate user registration on third-party sites without exposing the user's raw secrets, recovery materials, or issued credentials to the AI orchestration layer.
- The standard covers registration profiles, workflow orchestration boundaries, secret retrieval, credential write-back, and post-registration lifecycle handling.
- The standard does not replace browser automation, email handling, SMS handling, or anti-bot execution environments.
2. Roles and Trust Boundaries
Defines the actors in the registration hub model and what each actor may learn.
- User: the account owner whose onboarding profile, recovery data, and resulting credentials are being managed.
- Third-Party Integrator: the backend system that owns the CNothing client private key and decides which registration workflows are allowed to run.
- CNothing: the trust center that authenticates client requests, stores protected records, and returns encrypted data only to the authorized client backend.
- AI Agent: the orchestration layer that may decide next actions, navigate forms, and request protected records through CNothing-mediated flows, but MUST NOT possess decryptable user secrets.
- Execution Workers: browser, email, or verification workers that perform concrete actions under the policy decisions of the integrator.
3. Why CNothing Sits at the Center
Explains the operational value of CNothing in a registration control plane.
- CNothing gives the integrator a single protected place to register client identity, rotate keys, and bind every sensitive action to one-time challenge validation.
- CNothing lets AI agents request signup data without receiving long-lived plaintext secrets when the integrator uses private or blind record storage.
- CNothing centralizes post-registration credential storage so subsequent login, token refresh, and recovery operations can reuse the same protocol boundary.
- CNothing makes registration automation auditable because registration, refresh, read, save, and key-rotation events all pass through a common authenticated service contract.
4. Canonical Data Model
Defines the minimum logical objects expected in a conforming deployment.
4.1 Registration Profile
Represents the user-supplied onboarding material needed to complete signup.
- A registration profile SHOULD include the intended target site identifier, account identity hints, preferred channel bindings, and any required human-supplied enrollment fields.
- Sensitive fields such as passwords, recovery answers, API keys, or invite tokens SHOULD be stored using `savePrivateJson()` or `saveBlindJson()`.
4.2 Signup Target
Represents the external website or application being registered.
- A signup target SHOULD describe the registration entrypoint, required fields, expected verification channels, and any site-specific policy flags.
- A signup target MAY include reusable automation hints for browser or agent workers, but MUST NOT include user private key material.
4.3 Credential Bundle
Represents the result of a successful registration flow.
- A credential bundle SHOULD contain the issued username, email binding, password reference, recovery artifacts, API credentials, and session bootstrap material relevant to the target system.
- Credential bundles SHOULD be written back to CNothing immediately after successful registration so the integrator can continue lifecycle operations without replaying signup.
5. End-to-End Workflow
Specifies the recommended end-to-end orchestration flow for AI-assisted registration.
- The AI layer SHOULD receive only the minimum action-scoped data required to complete the current step.
- A conforming deployment SHOULD treat email codes, SMS codes, CAPTCHAs, and recovery paths as separate worker concerns, not as direct CNothing protocol features.
1. Third-party backend registers or refreshes its CNothing client identity.
2. Backend stores onboarding records in CNothing using blind or private mode.
3. AI agent begins a signup run against a third-party website.
4. When sensitive fields are needed, the agent requests the backend to fetch the protected record through CNothing.
5. The backend decrypts the CNothing response locally and returns only the minimum next-step data to the execution worker.
6. Verification artifacts and issued credentials are written back to CNothing after each milestone.
7. Final credential bundle is sealed and retained for future login, recovery, or rotation flows.
6. Recommended Storage Patterns
Defines how integrators should map onboarding artifacts into CNothing records.
- Use separate namespaces for signup profiles, active runs, credential bundles, and recovery records.
- Use `saveBlindJson()` for namespaces, keys, metadata, and values that should remain opaque to CNothing operators.
- Use `savePrivateJson()` when only the value and metadata need to be hidden from CNothing operators while namespace and key names remain operationally visible.
- Store run-local state independently from long-lived credentials so a failed workflow does not force credential record churn.
7. AI Agent Contract
Defines how the AI orchestration layer participates without becoming a secret holder.
- The AI agent MAY decide which workflow step should execute next.
- The AI agent MAY request that the third-party backend fetch protected records from CNothing.
- The AI agent MUST NOT be treated as a holder of client private keys, recovery bundles, or reusable login credentials.
- The AI agent SHOULD receive redacted or step-scoped values instead of full credential bundles whenever possible.
8. Post-Registration Lifecycle
Defines how CNothing continues to serve after account creation succeeds.
- The same CNothing client identity SHOULD be reused for subsequent login, password rotation, token refresh, and recovery workflows.
- Key rotation SHOULD be performed through the authenticated `rotateKey()` workflow instead of registering a replacement client identity.
- Integrators SHOULD preserve a durable mapping between their local user identity and the CNothing record set that stores the resulting credentials.
9. Security Requirements
Captures the minimum security posture required for a conforming registration hub deployment.
- Client private keys MUST remain under third-party backend control only.
- Blind mode SHOULD be the default for values whose names, metadata, and payloads are all sensitive.
- Execution logs SHOULD avoid storing raw user passwords, recovery codes, or decrypted credential bundles.
- Workers that touch verification channels SHOULD operate with the least privilege required for the current step.
- The integrator SHOULD maintain explicit operator approval or policy rules before an AI agent can initiate a new third-party signup run.
10. Conformance Checklist
Provides a quick checklist for independent implementers.
- Uses CNothing as the authenticated storage and retrieval boundary for signup-sensitive records.
- Keeps the client private key outside the AI orchestration layer.
- Stores onboarding secrets and resulting credentials in private or blind mode.
- Writes back issued credentials and recovery artifacts immediately after each successful step.
- Separates browser, email, SMS, and recovery workers from CNothing protocol responsibilities.
- Uses authenticated key rotation instead of replacing client identities when key material changes.