Agentic Interface Surfaces

Agentic Interface Surfaces (AIS) define the boundary through which an acting system encounters constraints, attempts effects, receives feedback, and earns recognition for progress. AIS shifts emphasis from instructions and narrative reports toward effect-bound evidence: a claimed action is not recognized as successful merely because an agent says it completed, a command returned zero, or one local test passed. The central rule is no admissible receipt, no recognized progress. Recognition requires an admitted action, an effect observation, evidence bound to the exact claim and environment, validation under a versioned rule, and a recognition decision with a declared scope. This rule does not make correctness inevitable. Effects may escape observation, evidence may be incomplete or forged, validators may encode the wrong target, and agents may optimize the receipt rather than the intended outcome. AIS complements Semantic Choke Point Binding: choke points identify where continuation can be constrained, while AIS defines what the agent sees, what evidence must be produced, and what feedback or next action follows. The paper specifies the architecture, its observability ceiling, and an evaluation program. It supplies no implementation or measured evidence.