Answers are grounded in this paper and cite their sources.
Protocol specification demonstrates an open governance architecture for AI agents, highlighting cryptographic auditability and enforced operational boundaries.
Agentic AI systems act — they call tools, spend money, message people, and change production state — yet the protocols connecting them standardise only capability. GRIP is a small, open protocol that makes agent work authorised before, bounded at an enforcing gate, and cryptographically attributable and integrity-verifiable afterwards, independent of model, framework or vendor: six signed JSON objects on one hash chain (Mandate, Grant, Receipt, Stop, Settlement, Context Envelope), a five-family act registry, a typed denial-reason registry, and four conformance levels. Status: Request for Comments. Proposed, not adopted; no implementers outside the authors. The specification artifacts — JSON Schemas, five registries, the eight-rule chain verifier, seven signed conformance vectors and the conformance runner — are served at grip.foxora.dev.
No takes yet. Share an insight, caveat, or question.
Parashar et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: