WebMEM Protocol Specification
Last Updated: 2025-07-28
Authors: David W. Bynon
Version: 1.0 Draft
What is the WebMEM Protocol?
The WebMEM Protocol is an open framework for publishing structured, machine-ingestible memory fragments—designed to be cited, retrieved, and trusted by AI systems.
Unlike traditional SEO or page-level metadata (e.g., Schema.org), WebMEM operates at the fragment level, enabling specific facts, glossary terms, explainers, and citations to be independently retrievable and trust-scored with provenance, context, and scoring logic.
Core Components
| Component | Description |
|---|---|
| SDT – Semantic Data Template | YAML-in-HTML and DATA-in-HTML templates embedded in pages using <template data-visibility-fragment>. Each fragment includes trust metadata, glossary alignment, and multi-format export compatibility. |
| GTP – Glossary Term Protocol | Defines the structure, versioning, and canonical publishing model for DefinedTerm and DefinedTermSet fragments with sameAs and provenance resolution. |
| ProvenanceMeta | W3C PROV-compliant metadata embedded at the fragment or digest level to assert source lineage, attribution, and dataset trust scope. |
| Digest Endpoint Exposure | Stable URIs expose digest fragments and full entities in multiple MIME formats (e.g., JSON-LD, TTL, MD, XML) using HTTP content negotiation. |
| AI Retrieval Feedback Loop | Optional mechanism for logging fragment retrieval events to reinforce visibility, track decay, and guide memory-first optimization. |
| SFI – Semantic Feedback Interface | RFC-007 spec enabling structured agent feedback to fragments or digests—supporting corrections, confidence scores, or clarification triggers. |
Lifecycle
- Authoring: Structured facts, definitions, and terms are authored as fragments.
- Digest Generation: Multi-format outputs are compiled as exportable semantic digests.
- Fragment Embedding: Memory objects are embedded in live HTML using inert containers.
- Endpoint Exposure: Each digest is made accessible via URI with content negotiation support.
- AI Retrieval: Agents retrieve fragments and optionally reinforce them via feedback.
- (Optional): Retrieval logs guide memory fitness dashboards and citation tuning.
Supported Formats
application/ld+json(JSON-LD)text/turtle(TTL)text/markdown(MD)application/xmltext/csvapplication/provenance+xml(PROV-O)
Protocol Modules
- SDT Specification – fragment format, attributes, and examples
- GTP Specification – term publishing and glossary resolution
- ProvenanceMeta – trust-layer metadata and validation
- Digest Endpoints – URI patterns and format negotiation
- AI Feedback Loop – retrieval logging + reinforcement
- SFI Specification – structured agent feedback and review scaffolds
WebMEM vs Schema.org
| Feature | Schema.org | WebMEM |
|---|---|---|
| Output Format | JSON-LD only | Multi-format (TTL, PROV, MD, etc.) |
| Scope | Page-level | Fragment-level |
| Binding | Loose (mainEntity) |
Precise (data-id, data-glossary-id) |
| Provenance | Optional/unsupported | Native W3C PROV support |
| Memory Optimization | ❌ | ✅ via AI Feedback Loop |
| Glossary Alignment | ❌ | ✅ via GTP + DefinedTerm |
Interoperability with MCP
| MCP Concept | WebMEM Complement |
|---|---|
tool.call() or RPC invocation |
Endpoint retrieves structured digest |
resource.type = digest |
WebMEM delivers multi-format entity memory objects |
| Retrieval + Reinforcement | WebMEM logs agent access and scores trust fitness |
| Citation Support | Fragment-level IDs + ProvenanceMeta allow exact trace |
| Structured Feedback | SFI enables POST-based agent feedback (e.g., /feedback/{data_id}) |
Status
The WebMEM Protocol is in open public draft. All RFCs, validator tools, and memory fragment templates are published at WebMem.com and versioned at github.com/WebMEM.
For contributions, technical alignment, or licensing questions, contact WebMEM.com.