MSP-1 - AI-friendly semantics for trusted information.
A reusable foundation for bringing MSP-1 awareness and functionality into publishing systems, development environments, and applications that consume information.
Status: Concept / Planning
Structural validation would remain separate from factual verification. Publisher declarations provide context; consuming applications remain responsible for assessing evidence and reaching conclusions.
Planned publishing integrations would help generate and maintain site and page declarations from publisher-managed information. Static sites, CMS platforms, and proprietary publishing tools could use the same foundation through interfaces suited to their architecture.
Common guidance and callable tools could help coding agents implement, inspect, update, and remove MSP-1 support. Codex, Claude Code, and Cursor are proposed integration targets; compatibility will be documented as individual integrations are developed and tested.
Repository integrations could support local declaration files, automated validation, and checks during development. Examples would show how to add MSP-1 handling to an existing project with a small, clearly defined integration surface.
Audit tools, knowledge graphs, and other information consumers could use declarations to retain identity, provenance, purpose, and interpretive context. Adapters would preserve the distinction between publisher assertions and independently established findings.
The proposed architecture separates core declaration handling from fetching, application workflows, and model-specific behavior. Processing a supplied declaration should be possible without a model, an account, or a live connection to an MSP-1 service.
A shared input/output contract and validation fixtures would support consistent behavior across interfaces. A command-line interface, language bindings, and optional agent-facing interfaces could be introduced as integrations demonstrate a need. Constructing declarations from known values would remain independent of optional AI-assisted drafting, which recommends publisher review.
Reversibility is a design requirement. Integrations should preserve source content and existing application structures, keep MSP-1 additions identifiable, and allow those additions to be disabled or removed cleanly.
When MSP-1 context is unavailable or support is disabled, the host should continue through its ordinary workflow. Reference implementations are intended to demonstrate installation, use, disabling, and removal as part of the same integration lifecycle.
This page defines the initial direction. No installable library or verified integration is being announced here. Package names, implementation languages, licensing, and release interfaces remain to be established.
The initial scope is a core implementation, a command-line interface, shared agent guidance, and one or two reference integrations. A proposed first candidate is an agent-based audit workflow; a different consumer, such as a knowledge graph project, would help test portability.
Source code, examples, and release documentation are intended to live in a dedicated repository under the MSP-1 Protocol GitHub organization. Repository and package links will be added here as those resources become available.