Concepts
IntentControl
IntentControl evaluates a proposed action against identity, authority, delegation, Intent, policy, current state and constraints, and returns a decision before the action has a consequence.
IntentControl is documented here ahead of release so integrators can design for it. Three separate answers today, shown below: commercially, it is coming soon in TrustOps and not yet on sale; signed-in operation: not yet available; there is no public decision API. Nothing is ENFORCED by ECZ-ID.
IntentControl today · measured 2026-10-10
- Commercial capability
- Coming soon · not yet on saleTrustOps lists every plan as coming soon. Nothing can be bought or provisioned yet.
- Signed-in operation
- Not yet availableThere is no signed-in operation until the service is confirmed live.
- Public developer interface
- Release-gatedThere is no public decision API, endpoint, credential format or SDK. Nothing is enforced by ECZ-ID.
Accountability Chain:Decision
What it evaluates
A proposed action reaches IntentControl from an enforcement point that sits in front of the action: an MCP gateway in front of tools, an HTTP gateway in front of an API, or an embedded check in your code. IntentControl runs separate checks, and passing one never stands in for another.
| Dimension | The question it answers |
|---|---|
| Identity | Is the actor a protected entity, and is its standing readable? |
| Trust state | Is the actor’s current state readable and acceptable? |
| Intent | Is there an Intent in force for this actor, and is this action inside it? |
| Authority | Was the Intent authorised by an accountable principal? |
| Delegation | Does every link in the delegation chain still hold, and narrow? |
| Time | Is the action inside the Intent’s window? |
| Policy | Does a published policy permit, forbid or require approval? |
| Constraints | Action, resource, destination, parameters, value, currency, data classes, rate and approval. |
The laws it holds
- Identity is not purpose, purpose is not authority, authority is not Intent, and Intent is not control.
- An allow means every configured check passed for this one action. It is not a statement that the actor or the action is safe, legal, correct or certain to succeed.
- A denial dominates every other outcome. Missing input never permits. An unreadable standing, or a receipt that cannot be written, never permits.
- Payment or binding is not authority. An action with no applicable published policy is refused.
- A delegated Intent only narrows, and delegation never multiplies a budget.
- A sandbox never grants production.
- Evaluated is not enforced. Enforced requires fresh mediation evidence from an inline enforcement point.
How it is designed to be deployed
- Hosted decisions
- The enforcement point asks ECZ-ID Core for each decision and authenticates with its own credential.
- Local and hybrid decisions
- The enforcement point decides from a policy bundle valid for at most 300 seconds; a staler bundle fails closed. Approvals always need Core.
- Inline mediation
- The point sits in the action path and can block. Only inline points can ever evidence enforcement.
- Advisory mediation
- The point returns the decision to its caller, who chooses. Advisory reports never make posture enforced.
- External identities
- An actor can be a Passport holder or an external identity: OIDC, SPIFFE, DID, OAuth client, X.509 or an API-key label.
Related
This page is explanation, not authority. Current state is only ever the live Resolver record. Live facts on this page were measured read-only on 2026-10-10.