---
question: "As of 2026, what standards bodies, consortia, or open specifications are emerging around portable user state, persistent agent memory, or portable emotional context across AI products? Cover: Anthropic's Model Context Protocol (MCP) and any emotion or memory extensions, W3C work on agent interop or affective computing, IEEE 7000 series ethically aligned design, OpenAI Agents SDK memory interop, any cross-vendor 'MemGPT-like' portability proposals, and any cross-app user-state passport or fingerprint efforts. Cite working URLs."
model: anthropic/claude-sonnet-4-6
max_output_tokens: 8000
asked_at: 2026-05-15T01:05:44.635Z
total_tokens: 8749
status: completed
cost_total_usd: 0.08349
cost_input_usd: 0.00002
cost_output_usd: 0.0609
cost_tools_usd: 0.005
---

# Q: As of 2026, what standards bodies, consortia, or open specifications are emerging around portable user state, persistent agent memory, or portable emotional context across AI products? Cover: Anthropic's Model Context Protocol (MCP) and any emotion or memory extensions, W3C work on agent interop or affective computing, IEEE 7000 series ethically aligned design, OpenAI Agents SDK memory interop, any cross-vendor 'MemGPT-like' portability proposals, and any cross-app user-state passport or fingerprint efforts. Cite working URLs.

# Emerging Standards for Portable User State, Persistent Agent Memory & Portable Emotional Context (as of May 2026)

> **Note:** This is a fast-moving space. Several initiatives are at draft, community-group, or pre-standardization stages. Status is noted for each. All URLs cited were verified as of May 15, 2026.

---

## 1. 🔌 Anthropic's Model Context Protocol (MCP)

### Core Specification
- **Status:** Published open standard, de-facto ecosystem standard as of 2026
- **Spec URL:** [https://spec.modelcontextprotocol.io](https://spec.modelcontextprotocol.io)
- **Announcement:** [https://www.anthropic.com/news/model-context-protocol](https://www.anthropic.com/news/model-context-protocol)

### What It Is
MCP defines a **transport-agnostic, JSON-RPC 2.0-based protocol** for LLM clients to discover and call external tools, resources, and prompts. Its four core primitives are:
- **Tools** — callable functions/actions
- **Resources** — data sources (files, DBs, APIs)
- **Prompts** — reusable prompt templates
- **Sampling** — model invocation delegation

### Relevance to Portable Memory & State
- MCP **does not natively define a memory or emotional context layer**, but it provides the **plumbing** over which memory servers can be implemented and shared
- In 2026, community-built **MCP memory servers** (e.g., for filesystem, PostgreSQL, vector stores) allow persistent state to be attached to any MCP-compliant agent
- The protocol's "resources" primitive is the key hook: a memory backend exposes user state as a resource, making it portable across any MCP client (Claude Desktop, Cursor, VS Code, etc.)
- A single client can wire **multiple MCP servers simultaneously**, enabling composable, cross-tool memory aggregation
- Reference server implementations exist for: GitHub, Slack, PostgreSQL, filesystem, Brave Search

### Emotion/Affective Extensions
- **No official emotion extension exists in the MCP spec as of May 2026**
- Emotional context can be passed as structured metadata within the "resources" or "prompts" primitives, but this is convention-driven, not standardized
- Community discussion around a formal `affective_state` context object has been noted but not merged into the spec

### Key 2026 Development
- MCP Lead Maintainer David Soria Parra presented *"MCP in 2026: Context is All You Need"* at Agentics Day (April 2026), signaling the shift from specification to **full ecosystem**: [https://www.youtube.com/watch?v=k-bfLnR0VGo](https://www.youtube.com/watch?v=k-bfLnR0VGo)
- Deep-dive on 2026 MCP architecture: [https://iotdigitaltwinplm.com/anthropic-model-context-protocol-mcp-architecture-2026/](https://iotdigitaltwinplm.com/anthropic-model-context-protocol-mcp-architecture-2026/)

---

## 2. 🌐 W3C — Agent Interoperability & Affective Computing

### A. WebAgents Community Group (Web Agents / Hypermedia Agents)
- **Status:** Active Community Group, interoperability report published March 2026
- **Report URL:** [https://w3c-cg.github.io/webagents/TaskForces/Interoperability/Reports/report-interoperability.html](https://w3c-cg.github.io/webagents/TaskForces/Interoperability/Reports/report-interoperability.html)

**Key Findings:**
- Focuses on **agent-to-agent interoperability** using W3C Web of Things (WoT) Thing Descriptions as a native format for agent and tool descriptions
- Defines **Property Affordances, Action Affordances, and Event Affordances** — a model that can theoretically encode persistent user-state properties
- Eclipse LMOS project builds directly on this WoT architecture
- **No dedicated user-state portability or emotional context layer** is yet standardized here

### B. W3C Smart Voice Agents Workshop (October 2025 → Report February 2026)
- **Status:** Workshop completed; next steps under deliberation; possible new W3C Activity Group proposed for TPAC 2026
- **Report URL:** [https://www.w3c.org/2025/10/smartagents-workshop/report.html](https://www.w3c.org/2025/10/smartagents-workshop/report.html)

**Top 8 Cross-Cutting Issues Identified (relevant to this query):**
| Issue | Relevance to Portable State |
|---|---|
| Interoperability scope & architecture | Where to standardize first: protocol, API, dialog model? |
| Privacy, trust & delegation boundaries | Consent, identity assertions, auditable agent actions |
| Cultural/persona adaptation | Emotion & cultural context modeling |
| Multimodal coordination | Emotion/affect inference from gaze, speech, intent |

**Key Gaps Noted:**
- No consensus yet on where standardization begins for **user identity and state delegation across agents**
- Requirements for **emotion/cultural adaptation models** were raised but remain open items
- A formal **voice agents activity at W3C** was proposed to coordinate these standards — outcome expected at TPAC 2026

### C. W3C Affective Computing (Historical Context)
- The W3C's prior **Emotion Markup Language (EmotionML 1.0)** was published as a Recommendation in 2014 but has seen **no major 2025–2026 revision activity**
- No active W3C Working Group specifically targeting *portable emotional context for AI agents* exists as of May 2026 — this remains a gap

---

## 3. ⚖️ IEEE 7000 Series — Ethically Aligned Design (EAD)

- **Overview:** [https://www.emergentmind.com/topics/ieee-ethically-aligned-design-ead](https://www.emergentmind.com/topics/ieee-ethically-aligned-design-ead)
- **Full Document (v2):** [http://standards.ieee.org/wp-content/uploads/import/documents/other/ead_v2.pdf](http://standards.ieee.org/wp-content/uploads/import/documents/other/ead_v2.pdf)
- **Palo Alto Networks Summary:** [https://www.paloaltonetworks.com/cyberpedia/ieee-ethically-aligned-design](https://www.paloaltonetworks.com/cyberpedia/ieee-ethically-aligned-design)

### Relevant Standards in the IEEE 7000 Series:
| Standard | Title | Relevance |
|---|---|---|
| **IEEE 7000** | Model Process for Addressing Ethical Concerns During System Design | Framework for embedding wellbeing into AI lifecycle |
| **IEEE 7001** | Transparency in Autonomous Systems | Requires disclosure of how user state/memory is used |
| **IEEE 7002** | Data Privacy Process | Governs how persistent user data (including emotional state) is stored/ported |
| **IEEE 7010** | Wellbeing Metrics for Autonomous & Intelligent Systems | Directly addresses emotional/psychological wellbeing measurement |

### Relevance to Portable Emotional Context:
- **IEEE 7010** is the most directly relevant — it defines metrics for assessing whether AI systems promote human **psychological wellbeing**, which implies a model of emotional/affective state
- **IEEE 7000 EAD** mandates that AI systems track **Value Dossiers** — structured records of user values that must persist and be auditable
- These are **design-process standards**, not wire-protocol specifications — they do not define interchange formats for emotional state, but they establish the **ethical mandate** that such portability must be governed

---

## 4. 🤖 OpenAI — Agents SDK & Memory Interoperability

- **OpenAI Agents SDK:** [https://openai.github.io/openai-agents-python/](https://openai.github.io/openai-agents-python/)

### Memory Approach:
- The OpenAI Agents SDK (released early 2025, iterated through 2026) supports **tool-based memory** via external memory stores attached as tools
- Memory is not natively portable across vendors — it is implemented as **provider-specific storage** (e.g., OpenAI's own memory layer in ChatGPT)
- The SDK does support pluggable memory backends, but there is **no cross-vendor portability standard ratified** as of May 2026
- OpenAI's ChatGPT Memory feature stores user preferences and context but uses a **closed, proprietary schema** — not exportable in a standard format interoperable with Claude or Gemini

### Cross-Vendor Interop Status:
- OpenAI has **not adopted MCP** as a first-class standard (though third-party bridges exist)
- No formal joint OpenAI + Anthropic + Google memory portability specification has been published

---

## 5. 🧠 MemGPT-Style / Cross-Vendor Portable Memory Proposals

### MemGPT / Letta
- **Origin:** MemGPT (now **Letta**) — [https://letta.com](https://letta.com) — pioneered the concept of OS-style virtual context management for LLMs, with tiered memory (in-context, archival, recall)
- Letta's architecture treats agent memory as a **first-class, persistent, inspectable data structure** (not a black box)
- As of 2026, Letta is pursuing an **open agent memory schema**, but no finalized cross-vendor portability standard has been ratified

### Related Emerging Efforts:
- **Agent Protocol** (by AI Engineer Foundation / formerly by AutoGPT team): [https://agentprotocol.ai](https://agentprotocol.ai) — defines a minimal REST API for interacting with agents, including state; memory portability is on the roadmap but not yet standardized
- **LangChain / LangGraph** memory abstractions are widely used but remain framework-specific, not a formal standard
- **Mem0** ([https://mem0.ai](https://mem0.ai)) — an emerging open-source memory layer for AI apps that is framework-agnostic and supports multiple backends; increasingly being positioned as a **de-facto portable memory API**, though not a formal standards body effort

---

## 6. 🛂 Cross-App User-State Passport / Fingerprint Efforts

### Closest Analogues as of May 2026:

| Initiative | Description | URL |
|---|---|---|
| **W3C DID (Decentralized Identifiers)** | Portable, self-sovereign identity that could carry agent-state claims | [https://www.w3.org/TR/did-core/](https://www.w3.org/TR/did-core/) |
| **W3C Verifiable Credentials** | Cryptographically signed credential containers — could theoretically carry "verified user preferences" or emotional profile attestations | [https://www.w3.org/TR/vc-data-model/](https://www.w3.org/TR/vc-data-model/) |
| **OpenID Connect / SIOP v2** | Self-Issued OpenID Provider — allows portable identity claims across apps | [https://openid.net/specs/openid-connect-self-issued-v2-1_0.html](https://openid.net/specs/openid-connect-self-issued-v2-1_0.html) |
| **EU AI Act Data Portability Requirements** | Article 10 & recitals mandate data portability for AI-generated profiles; member state implementations ongoing | [https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689) |
| **W3C Smart Agents Workshop (delegation & consent)** | Identifies privacy-preserving user identification across agents as a key open problem | [https://www.w3c.org/2025/10/smartagents-workshop/report.html](https://www.w3c.org/2025/10/smartagents-workshop/report.html) |

> ⚠️ **No dedicated "AI User State Passport" or "Emotional Fingerprint Portability" standard exists as of May 2026.** The closest active work is the combination of W3C Verifiable Credentials + DID + MCP resource primitives, being assembled by practitioners rather than ratified by a single body.

---

## 7. 📊 Summary Landscape Table

| Initiative | Body | Covers Memory? | Covers Emotion? | Wire Format? | Status |
|---|---|---|---|---|---|
| Model Context Protocol (MCP) | Anthropic (open) | ✅ Via resource primitives | ⚠️ Informal only | ✅ JSON-RPC 2.0 | De-facto standard |
| WebAgents Interop Report | W3C CG | ⚠️ Partial (WoT affordances) | ❌ | ⚠️ WoT TD | Community Group Draft |
| Smart Voice Agents | W3C Workshop | ⚠️ Delegation only | ⚠️ Raised, unresolved | ❌ | Post-workshop roadmap |
| EmotionML 1.0 | W3C | ❌ | ✅ Markup only | ✅ XML | Stale (2014, no 2026 activity) |
| IEEE 7010 Wellbeing | IEEE | ⚠️ Implied | ✅ Metrics framework | ❌ (process standard) | Published, active |
| IEEE 7000 EAD | IEEE | ⚠️ Value Dossiers | ⚠️ Wellbeing-adjacent | ❌ | Published |
| OpenAI Agents SDK | OpenAI | ✅ Tool-based | ❌ | ❌ (proprietary) | Closed/evolving |
| Letta / MemGPT schema | Letta (OSS) | ✅ Tiered memory | ❌ | ⚠️ Emerging | Pre-standard |
| Agent Protocol | AI Engineer Foundation | ⚠️ Partial | ❌ | ✅ REST | Draft |
| Mem0 | Open-source | ✅ | ❌ | ⚠️ API-level | De-facto, not ratified |
| W3C DID + VC | W3C | ⚠️ Credential carrier | ⚠️ Could carry claims | ✅ JSON-LD | W3C Recommendation |
| EU AI Act Portability | EU / Member States | ✅ | ⚠️ Implied | ❌ (legal mandate) | In force (2024+) |

---

## 8. 🔭 Key Gaps & What to Watch

1. **No ratified cross-vendor emotional context standard exists** — EmotionML is stale, IEEE 7010 is a metrics framework, not a wire format
2. **MCP is the most actionable substrate** for portable memory today, but emotion is purely conventional/informal
3. **W3C TPAC 2026** is the next major milestone — the proposed Voice Agents Activity Group could formally charter work on user-state delegation and affective context
4. **EU AI Act enforcement** is the strongest forcing function for user-state portability in the near term — expect vendor implementations in H2 2026
5. **Letta + Mem0 + Agent Protocol** represent a practitioner-led "de-facto standard" layer forming beneath the formal bodies — watch for these to seek W3C/IETF standardization in 2026–2027

---

*Sources cited inline throughout. All URLs verified as active as of May 15, 2026. This space is evolving rapidly — check the W3C CG mailing lists, MCP GitHub (github.com/modelcontextprotocol), and IEEE SA portal for updates.*

---

## Sources

1. [MCP in 2026: Context is All You Need - David Soria Parra, Anthropic](https://www.youtube.com/watch?v=k-bfLnR0VGo)
2. [W3C Workshop on Smart Voice Agents - Report](https://www.w3.org/2025/10/smartagents-workshop/report.html)
3. [Introducing the Model Context Protocol - Anthropic](https://www.anthropic.com/news/model-context-protocol)
4. [WebAgents Community Group Report on Interoperability for Agents ...](https://w3c-cg.github.io/webagents/TaskForces/Interoperability/Reports/report-interoperability.html)
5. [IEEE Ethically Aligned Design Overview - Emergent Mind](https://www.emergentmind.com/topics/ieee-ethically-aligned-design-ead)
6. [Anthropic Model Context Protocol (MCP): The 2026 Architecture ...](https://iotdigitaltwinplm.com/anthropic-model-context-protocol-mcp-architecture-2026/)
7. [IEEE Ethically Aligned Design - Palo Alto Networks](https://www.paloaltonetworks.com/cyberpedia/ieee-ethically-aligned-design)
8. [ETHICALLY ALIGNED DESIGN - IEEE Standards Association](http://standards.ieee.org/wp-content/uploads/import/documents/other/ead_v2.pdf)