MSP-1 - AI-friendly semantics for trusted information.
Extensions is the Labs area for proposed MSP-1-compatible structures that may expand what the ecosystem can declare without expanding the minimal core protocol itself.
MSP-1 is intentionally small. Extensions provide a controlled way to explore additional semantic patterns while keeping the core protocol stable, readable, and broadly implementable.
A good extension should add useful context without creating dependency, enforcement, or protocol bloat.
Proposed extensions should follow the same discipline as MSP-1 itself: clear scope, conservative claims, graceful degradation, and compatibility with systems that do not recognize the extension.
Extension concepts for or in development.
Provide an optional, declarative layer for identifying brand authority, origin, affiliation, evidence boundaries, and official representation across agentic systems. Its purpose would not be to optimize recommendation outcomes, assert market leadership, or create a proprietary scoring framework, but to help machines distinguish official brand context from derivative, outdated, unaffiliated, or unsupported representations.
Introduces an explicit, declarative asset registry that allows publishers to identify the protocols, tools, APIs, knowledge bases, identity anchors, structured data surfaces, and functional capabilities available across a site, page, component, or endpoint.
ADE provides agents with an architectural map of what assets exist, where they are located, what tasks and contexts they support, and how they relate to one another. This helps agents identify a relevant working set before parsing individual payloads, activating runtime environments, or inferring functionality from page structure alone.
ADE remains strictly declarative. It does not invoke tools, authorize actions, define protocol mechanics, or require agents to follow a prescribed path. It gives agents a clearer starting point while leaving execution to the appropriate native system.
A proposed structure for declaring task state, intent, context, and interpretive posture when one agent hands work to another agent, model, tool, or workflow stage.
A proposed structure for preserving milestone-level task context so agents can retain meaning, reduce drift, and avoid carrying unnecessary conversational or workflow history.
A proposed structure for declaring the origin, edit history, AI involvement, or creative lineage of media assets without turning MSP-1 into an enforcement or rights-management system.
A proposed structure for declaring the limits, assumptions, uncertainty, recency, or domain boundaries of a model output, document, dataset, or agent-generated artifact.
Extensions should mature gradually. An idea may begin as a concept, move into a draft structure, become testable through experiments, and only later be considered for broader documentation or tooling support.
An extension is not a request to make the core protocol larger. Labs extensions exist precisely because many useful ideas should remain optional, scoped, and experimental.
This separation allows MSP-1 to evolve around a stable center rather than absorbing every possible use case into its foundation.
Extension work connects naturally to concepts, experiments, evaluation, and developer guidance.