Skip to content

How It Works

Define intent. Interpret it. Verify it. Grant authority. Execute.

Intended sits between an agent deciding to act and the action actually happening. It verifies that the action matches what was intended, grants authority cryptographically, and records every decision in an immutable, independently verifiable audit chain.

Verification chain

Selected step

Open Intent Layer: define intent

The Open Intent Layer provides the shared language for what agent actions are intended to do across tools, systems, and workflows.

Layer definitions

Open Intent Layer

Define intent

The Open Intent Layer defines what agent actions are intended to do in a shared, open language.

This is what makes raw actions legible across tools, runtimes, and enterprise systems.

Large Intent Model (LIM)

Interpret intent

The Large Intent Model interprets the action in context so the system understands what the agent means, not just what it called.

This is how Intended reasons across domains and systems without collapsing everything into low-level API traces.

Intent Object

Structure intent

The Intent Object is the structured representation of what is supposed to happen and what the verification layer will evaluate.

This is the concrete unit of verification, replay, explanation, and evidence.

Enterprise Capability Engine (ECE)

Verify enterprise context

The Enterprise Capability Engine maps the Intent Object to enterprise capabilities and validates the operational context.

This is where technical intent becomes enterprise-legible and capability-aware.

Authority Engine

Make the decision

The Authority Engine evaluates policy against verified intent and returns ALLOW, DENY, or ESCALATE.

Policy is a component here, not the product category. The category is intent verification.

Authority Token + Execution + Audit

Grant authority and enforce it

Approved actions receive an Authority Token. Execution validates that token. Audit records the full chain as immutable evidence.

This is what makes the system enforceable, provable, and exportable.

What Intended is not

Intended is not a policy engine, an observability dashboard, a guardrail, or a filter. It is the verification and authority layer between agent intent and runtime execution.

Enforcement invariant

Execution validates the Authority Token at the boundary. If the token is missing, invalid, expired, or reused, execution is blocked.

No Token, No Action.

Next step

See the verification model in your own stack.