Portable
Switch between AI providers without losing accumulated context. Your data follows you, not your vendor.
Draft Specification v0.2
Your AI Agent's Memory,
Your AI Agent's Memory,
Under Your Control
An open specification for portable, private AI agent data. Store memory, capabilities, and state in user-controlled Solid Pods.
The Problem
Agent data is siloed
AI agents accumulate valuable context—your preferences, interaction history, learned behaviors. Today, this data is locked in vendor-specific formats within proprietary infrastructure.
Vendor Lock-in
Switching providers means starting from scratch
No Audit Trail
Users can't review what agents have learned
No Sharing
Context can't move between different agents
Our Approach
Design principles
Private
Data encrypted by default. You control who sees what through granular, explicit permissions.
Interoperable
Built on W3C standards: Solid Protocol, RDF, DIDs. Works with existing web infrastructure.
Foundation
Built on open standards
Documentation
Get started
Full Specification
Pod structure, data formats, access control, conformance requirements, and security considerations.
RDF Vocabulary
Classes, properties, and SHACL shapes. Extends FOAF and PROV-O.
W3C Position Paper
Proposal to the AI Agent Protocol Community Group.
Interoperability
Integration with ANP, A2A, MCP, and other protocols.
Help shape the future of agent data
The specification is in draft status. Implementation feedback and contributions are welcome.