A simple protocol that keeps humans in control—by forcing AI to pause when meaning is unclear.
Modern AI systems don't wait for commands — they initiate actions, trigger workflows, and make decisions. Without a source of human-defined intention, they improvise:
That's not intelligence.
That's automation without direction.
The Human Agency Protocol installs the missing layer.
Integrated directly into AI platforms, HAP forces systems to pause when meaning, purpose, or intention is unclear — and resume only once humans decide what the action is for.
With HAP, your AI acts on human intention, not assumptions.
The system detects ambiguity, drift, or a skipped stage.
A context-appropriate question is triggered (an Inquiry Blueprint).
Only after the human resolves the checkpoint can the AI continue.
These checkpoints are enforced. They cannot be bypassed.
Human thinking has two rhythms. HAP protects both.
For alignment, planning, and execution.
For exploration, insight, and creative depth.
Both modes follow the same rule:
AI cannot advance when orientation is unclear.
All productive human inquiry follows the same structure:
Are we talking about the same thing?
Why does this matter?
What will we do first?
How do we execute?
HAP enforces progress through these stages.
When any stage is unresolved, the system must pause and ask.
AI must follow the ladder in the correct order (Convergent) or controlled cycles (Reflective).
No jumps. No shortcuts. No inference.
When meaning, purpose, or intention is unclear, AI must stop and ask.
Explicit confirmation required.
All downstream actions (publishing, deploying, triggering systems) require resolved stages.
Higher risk → higher required stage.
Convergent for decisions; Reflective for depth.
Both obey stop → ask → proceed.
Only structural signals leave the device.
Never content. Never transcripts.
Questions follow open, shared Inquiry Blueprints.
AI chooses wording; the protocol enforces timing.
Every checkpoint and action can be cryptographically proven (HAP Envelope).
Proof, not trust.
Qualified Service Providers enforce the rules without owning data.
No central authority. No extraction.
Automation accelerates. Human intention becomes scarce.
HAP ensures:
The right question, asked at the right time, keeps humans in the loop — by design, not by hope.
How agency is described, measured, and exchanged
Tools for building AI that obeys the protocol
Infrastructure enforcing compliance at scale
Verifiable, federated oversight without centralization