open-memory-object

Open Memory Object (OMO) - Universal standard for AI memory interoperability with built-in safety, privacy, and consent management. Enabling seamless memory portability across platforms, databases, and AI systems while prioritizing user sovereignty and frontier AI safety.

Open Memory Object (OMO)

Universal Standard for AI Memory Interoperability

🚨 Status: This consortium is actively forming. The specification has not yet been ratified by founding members.

Why OMO?

LLM-powered agents and AI applications depend on long-term memory for personalized experiences. Today, that memory is fragmented across proprietary formats, creating:

• Platform Lock-in - OpenAI, Anthropic, Google store memory in closed schemas
• Integration Tax - Developers build bespoke adapters for each memory platform
• Data Sovereignty Loss - Users can't export, audit, or delete their memories
• Safety Gaps - Inconsistent consent, risk assessment, and content filtering
• Innovation Drag - Startups duplicate infrastructure instead of building features

OMO solves this with a universal, safety-first standard for AI memory objects.

Core Principles

🛡️ Safety First

Every memory object carries built-in safety metadata, enabling automatic risk assessment and content filtering before reaching AI models.

👤 User Sovereignty

Memory belongs to users, not platforms. All operations must respect explicit consent and provide "forget-me" controls.

🔒 Privacy by Design

Consent tracking is mandatory. Every memory object explicitly records how consent was obtained and permissions granted.

🔗 Universal Interoperability

One standard across memory platforms, databases, cloud providers, and AI systems. All specifications, reference implementations, and governance processes are publicly available under permissive licenses.

⚡ Developer Simplicity

Single integration point reduces complexity and accelerates innovation.

Schema Overview

{
  "id": "mem_abc123",
  "createdAt": "2025-06-26T10:30:00Z",
  "type": "text",
  "content": "User prefers morning meetings",
  "consent": "explicit",
  "risk": "none",
  "topics": ["scheduling", "preferences"],
  "acl": {
    "read": ["user:alice", "app:calendar"],
    "write": ["user:alice"]
  },
  "ext": {
    "papr:relationships": [...],
    "qdrant:vector": [0.1, 0.2, ...]
  }
}

Participate & Commit (Organizations)

Help shape OMO by committing your organization’s participation. Add your org via Pull Request by appending a row to the table below, or open an issue titled "Organization Commitment" and we’ll add it for you.

Organization	Appointee (name/title)	Commitment Level	Contribution Notes	Workshop Interest	CLA	Date
Papr	Shawkat Kabbara / Founder & CEO	Founding Contributor	Lead a working group & contribute code	Yes	TBA	09/02/2025
Georgia Tech	Thomas Conte / Associate Dean for Research; Professor, Joint with ECE; Adjunct Faculty SCI	Founding Contributor	Willing to host workshop (venue)	Yes (Host)	TBA	09/02/2025
Neo4j	TBA	Founding Contributor	Willing to host our first workshop (venue)	Yes (Host)	TBA	09/04/2025
Qdrant	TBA			Yes	TBA
Exa	TBA			Yes	TBA

Notes:

• Please keep entries alphabetical by organization.
• "CLA" indicates whether your corporate CLA is signed or in progress.

Commitment Levels & Incentives

• Participant: Attend workshops and discussions; send at least one representative.
  • Incentives: Public listing, early workshop materials, recognition in recap post.
• Contributing Member: Pledge engineering time (e.g., 10–40 hours/month) or implementation work toward reference SDKs/spec.
  • Incentives: Logo on site, WG voting within working groups, priority review of RFCs.
• Founding Contributor: Lead or co-lead a working group, contribute substantial reference code, or provide critical infrastructure (compute, testing, CI).
  • Incentives: Eligibility for formation committee seats, co-authorship on initial spec drafts, prominent recognition.
• Sponsor: Fund events, documentation sprints, or certification infra; or provide venue/credits.
  • Incentives: Sponsor placement, stage mentions, priority booth/table at events.

Nudges we use to encourage contributions:

• Public leaderboard: We sort by commitment level and date. Earlier and higher commitments appear higher.
• Matching pledges: If you match another org’s pledge (time/infra), you’ll be co-highlighted in the next update.
• Milestone unlocks: When N orgs pledge at a level, we unlock perks (e.g., early certification beta).

See governance alignment on tiers and process: governance/GOVERNANCE.md#membership-tiers and governance/GOVERNANCE.md#participation--commitments-process.

Open Memory Workshop (Hosted by Neo4j)

• When/Where: TBA (date), TBA (location)
• Format: Problem-first working sessions to inform domains and roles; lightning talks; WG breakouts
• How to Join: Add your org to the table above with "Workshop Interest: Yes" and we’ll share details.
• Code of Conduct & Safety: We follow OMO safety/privacy guidelines.