2AYE Authorize
Does this exact action satisfy policy?
Deterministic evaluation and short-lived, single-use grants bound to material parameters.
Enterprise AI securityAI agents can act.
AI agents can act.
Make sure they have authority.
Verify identities, govern consequential actions, require exact human approval, and preserve evidence of execution.
Your agents. Your systems. An explicit authority boundary.Keep your existing identity provider, models and business systems. Start with one protected action and measurable acceptance criteria.
Human intent.
At the heart of every action.
At the heart of every action.
BUILT AROUND YOUR PEOPLEMore confidence.
Illustrative journeys. Availability depends on the integration and deployment; these selections do not execute an action.
More confidence.
More possibility.
Explore how your team fits into the authority loop.
Give teams room to move. Keep people in control.
Define what an agent may do, who owns the decision, and when a person must review it.
- Name an accountable owner
- Set purpose and limits
- Review consequential exceptions
People, devices, applications and AI agents
One authority platform. Six connected products.
Connect identity, accountable ownership, bounded authority, human review, risk context and evidence. Each capability keeps its own implementation, deployment and provider requirements.
2AYE Auth
Who is requesting access?
Authentication, step-up, and device context support the action-control boundary.
2AYE Agent
Which agent may act, and for whom?
Accountable ownership, scoped authority, and delegation that may only narrow.
2AYE Evidence
What was authorized, and what happened?
Correlate decisions, grants, destination outcomes, execution receipts, and hash-linked audit.
2AYE Approve
Who must approve this pending action?
Human decisions bound to an exact action, with enrolled-device signing and explicit expiry.
2AYE Risk
Which context requires intervention?
Contextual signals and rule-based policy inform access decisions and bounded response.
Explore products and capabilitiesCheck capability readinessRun the developer evaluation
Explore the security boundary
See what gets through.
And what never should.
Choose a request. Follow its path from human intent to the protected resource, and see which control must stop it.
Agent proposesAuthority stays outside the model
Model and tool outputUNTRUSTED INPUT
- 01Human owner
Accountable identity
Awaiting scenario - 02Signed intent
Explicit action and scope
Awaiting scenario - 03Policy check
Deterministic constraints
Awaiting scenario - 04Bound grant
Short-lived authority
Awaiting scenario - 05Protected resource
Verify, then redeem once
Awaiting scenario - 06Effect evidence
Verify the actual outcome
Awaiting scenario
Enforce at the resource that can perform the action
The proposed resource is outside the signed contract. Deterministic scope checks refuse it, whatever the model was told.
Inspect the security modelInteractive explanation using synthetic requests. It does not call the authorization API, issue a grant or execute an action. Explore real signature verification below.
From authorization to evidence
One action. One enforceable lifecycle.
Follow the boundary a consequential action must cross. Each stage exposes what is checked, what can stop execution, and what evidence survives afterward.
- 01One runtime path
- 02Fail closed
- 03Exact parameters
- 04Single use
- 05Hash-linked evidence
SIGNED SCOPE
Give the action an accountable owner and explicit limits.
Purpose, resource, version, expiry and delegation become signed material. A material change needs a new version and signature.
The contract is authority. An instruction from a model is not.
Inspect the enforcement contract Illustrative control outline
AUTHORIZATION CONTRACT
Accountable owner: required
Action: credential.rotate
Resource: service-key
Version and expiry: explicit
Delegation: may only narrowTrust & verification
See the evidence.
Know its boundaries.
Published formats and reference checks make the control model inspectable. Your deployment, provider and protected resource still need their own acceptance evidence.
Reference controls exercised
Results from the published reference verification summary.
Independent receiver validations
Independent interoperability remains an open acceptance gate.
Workflow before expansion
Agree exact limits, refusal cases, key custody and observable outcomes for your pilot.
Try real cryptographic verification
Change the receipt.
Watch the signature fail.
Verify published fixtures on your own machine, then change the amount without re-signing it. No account or live environment is needed.
2AYE Authenticator · within 2AYE Auth
Your identity.
Your approvals.
Your control.
Keep everyday verification codes on your device. When an organization connects you, review the requested action before deciding.
Everyday account protection
Offline authenticator codes and reviewed imports. Organization enrollment is a separate, invitation-based step.
A deliberate human decision
Review the request, its scope and expiry. An enrolled device signs a decision; approval never overrides a hard policy denial.
Reference application. Public App Store and Google Play distribution are not yet available. Complete device and runtime-approval acceptance remains open.
Rendered from the application’s Flutter widgets with synthetic fixture data and Android fonts. These are UI renders, not a live account, executed action or physical-device verification.
Enterprise coverage
Cover identity, access, agents, and connected enforcement.
Use one security model across employees, customers, autonomous agents, protected services, infrastructure, and MSP-operated tenants.
Coverage 01 / 05AI agents
Govern autonomous action
Bind agent identity and tool use to signed purpose, policy constraints, approvals, and execution evidence.
- 01Agent and tool identity
- 02Intent-bound authorization
- 03Human approval gates
Move from architecture to a governed workflow.
Use the documentation and quickstarts to identify the right integration path before production enforcement.
What feeds the decision
Explore the controls.
See how they connect.
Choose a product domain. Follow its path from identity and context to deterministic decisions and execution evidence. Open a capability to review its scope, or check deployment readiness below.
Selected signalAgent intent
01Observe
Identity and context
02Decide
Deterministic policy
03Authorize
Bound authority
04Prove
Execution evidence
2AYE Agent
Carry identity control into autonomous action.
Agent identity, signed intent, runtime authorization, and policy enforcement.From understanding to adoption
A clear next step at every stage.
- 01
Explore the model
See allow, review and refusal behavior in an illustrative browser demonstration.
Open the walkthrough - 02
Verify public artifacts
Run published verification material. Source access and production integration are separate evaluation steps.
Run the developer evaluation - 03
Scope your pilot
Choose one action, a protected resource, accountable owners and measurable outcomes.
Discuss an enterprise pilot - 04
Validate your deployment
Review key custody, exact-action enforcement, replay refusal, recovery and independent effect evidence.
Review acceptance requirements
Your next step
Bring one workflow.
Define what must never happen.
Agree the protected resource, exact action, human reviewers and measurable refusal and recovery cases. Expand when the evidence supports it.