Loyal Agents

Get Involved

Put Our Work Into Practice Early

Early adopters put our standards and research to work in real products and services, and tell us what works. Their experience shapes each version of the Human Context Protocol (HCP) as it develops, and grounds our research in real deployments.

A participant speaking during a working session at the workshop.
A working session at Portable Memory + HCP: Global Portable Memory Workshop, October 2026

The standard

Built to Fit the Tools You Already Use

HCP is an open standard that lets people hold their own preferences and context and decide which agents may use them. Version 2 organizes the standard into a small core, bindings for HTTP and the Model Context Protocol (MCP), and profiles for specific settings such as health, education and delegated authorization.

Learn more about HCP

The research

Research You Can Build With

Long-context memory

MemoryArena, presented at ICML 2026, tests whether agents can use what they remember across sessions to complete later tasks. Our team continues to publish on how agents remember and use personal context.

Explore MemoryArena

Safe multiagent systems

The System Prompt Index is an open library of more than 1,000 system prompts from over 400 AI products, each audited against AISPA, a user-centric assurance standard for the hidden instructions that shape an AI's behavior.

Explore the System Prompt Index

Early adopter program

What Early Adopters Get

  • Access to draft specifications as they develop.
  • Working sessions with the specification editors.
  • A direct say in the design: problems you find feed into the next draft.
  • An open-source reference implementation to build and test against.
  • Opportunities to collaborate on research in long-context memory and safe multiagent systems.

What We Ask

  • Implement part of a draft specification, or apply our benchmarks and tools, in a real product or service.
  • Report gaps, ambiguities and problems as you find them.
  • Share what you learn, so our standards and research work for others.

Building now

Who's Building Now

Reference partners Turn.io and the Learning Economy Foundation are testing the specification across multilingual, low-connectivity and education use cases. We're adding early adopters now.

Get involved

Tell Us What You're Building

Whether you run a consumer service, a learning platform or an agent product, we'd like to hear how our standards and research could fit.

Contact the TeamView the reference implementation