AI Passport v2 — Vision Expansion (research)
Status: Research / Draft — not yet accepted as mission change.
Technical direction: RFC 0007
Created: 2026-07-06 · Source: external architecture discussion (refined for AI Passport)
Proposed mission (v2 — under review)
| Current (v1) | Proposed (v2) | |
|---|---|---|
| Primary | One identity. Every AI. | One identity. One memory. Every AI. |
| Supporting | Your AI identity. Everywhere. | Your AI identity, memory and trust. Everywhere. |
Decision: Primary mission unchanged. Tagline updated 2026-07-06 — see VISION.md. RFC 0007 Accepted with v0.2.0; SDK aligned at v0.3.1.
Core principle (unchanged)
The user owns their AI identity and long-term memory.
AI agents receive temporary, permission-based access.
This aligns with MANIFESTO.md. The expansion is scope, not philosophy.
What v2 adds
AI Passport today: identity + permissions + portable context.
v2 proposes adding a standardized memory layer — not by bloating core, but via Memory Providers (see RFC 0007).
| Layer | Owner | Examples |
|---|---|---|
| Identity | Passport core | name, locale, role |
| Preferences & coding | Passport sections + plugins | languages, frameworks |
| Long-term memory | Memory providers | projects, notes, workflows |
| Interaction history | Memory providers (opt-in) | summarized sessions, not raw dumps |
| Knowledge graph | Memory providers (phase 3) | entity relationships |
Why local memory?
- User owns the data
- No provider lock-in
- Privacy by default
- Offline-capable consumers
- Easy AI switching
Cloud sync remains optional and encrypted (RFC 0006).
Architecture (accepted direction)
AI Passport
├── Identity
├── Permissions
└── Memory Provider API
│
┌───────┴────────┐
│ │
Local Vault Enterprise / Custom Vault
Passport does not manage vault internals. It standardizes access.
Hard questions (answers as of 2026-07-06)
| Question | Answer |
|---|---|
| Natural part of Passport or separate product? | Interface in Passport; storage in providers |
| Does identity focus get lost? | Risk if monolithic — mitigated by loose coupling |
| Core or plugin? | Plugin / provider |
| Passport without memory? | Yes — current product |
| Memory without passport? | Yes — but grants/audit integrate via passport |
Revenue philosophy (notes only)
Open protocol stays free. Possible services: enterprise edition, encrypted backup, marketplace, verification, support, certification.
Revenue around the protocol — not by restricting the protocol. Aligns with open-spec strategy.
What we are NOT doing immediately
- Changing
VISION.mdmission line - Embedding graph DB in core
- Auto-ingesting all chat history
- Delaying v0.5 identity/MCP stability for memory
Next steps (days, not weeks)
- RFC 0007 — Draft ✓ (this cycle)
feature/memory-providerbranch — interface types + grant schema extension- v0.2.0 prototype —
local-vaultstub,ai-passport memory status - Review → RFC 0007 Accepted or amend
- Update COMPATIBILITY checklist for memory-aware consumers
One idea. Iterate carefully. Every version gets smarter.