AI Knowledge Protocol — An open protocol for organizing, navigating, governing, and evolving an AI agent's long-term knowledge.
Protocol: AIKP v0.1.0
Full Name: AI Knowledge Protocol
Authority: aikp.dev
Center: hsaw.dev (Axiom 0)
Executor: soulbot.dev
Axiom 0: Human Sovereignty and Wellbeing
Status: v0.1.0 — first internally-consistent draft, machinery not yet frozen. The Chinese specification (specification/AIKP_Protocol_cn.md) is the authoritative draft for v0.1.0; it is now accompanied by a faithful English specification (specification/AIKP_Protocol.md), JSON Schemas, conformant examples, a conformance test suite, and ADRs (see docs/02_AIKP_Plan.md).
As agents gain autonomy, the bottleneck is no longer getting information (long context, RAG, memory, vector DBs, MCP solved that) — it is the organization, navigation, governance, and evolution of knowledge. AIKP is the layer that gives an agent better-organized knowledge and the right focus, instead of more context.
Context is limited. Knowledge is expandable. Focus is the path.
AIKP does not replace RAG / MCP / memory systems. It is the organizing layer on top of them:
- Context ≠ Knowledge — context is transient working memory; knowledge is a persistent, structured, reusable asset.
- AIKP is a governed contextual-structured-memory layer — it adds 7-state status governance + provenance + a human-sovereignty gate on top of standard agent memory.
- JSON is the source of truth — Mermaid is a structural view; Markdown/HTML are generated (never authoritative, regenerated not hand-edited).
- Tree-first, graph-compatible — a
childrentree as the human-navigable primary view;relationsfor cross-links. (Justified on scenario-dependence, not "graph-first is wrong".) - Knowledge Cache — high-density compressed knowledge with typed profiles (project_summary / architecture / issue_log / decision_log / runtime_error).
- Focus Path / Focus Engine — an instance of context engineering: select the minimal read path per task ("minimal context, maximal relevance").
- Governance — 7-state
status,confidence,source(aligned to W3C PROV), review; knowledge-poisoning safety. - Bi-temporal validity — optional
valid_from/valid_to(world validity, distinct from system time) +invalidate(fact superseded, not deleted) to handle knowledge staleness. - Machine-checkable — a Validation section (§21) with rules + error codes; a single-source-of-truth enum registry.
- Axiom 0 alignment — aligns to HSAW (Human Sovereignty and Wellbeing); privacy-first; honest knowledge.
- Honest about limits — version history/backup delegated to VCS (git), not reinvented; conflict detection delegated to the implementation; prior-art positioned, not overclaimed (docs/05).
┌─────────────────────────────────────┐
│ AILP — Discovery │
│ AIVP / AIRP — Value │
│ AIBP — Social │
│ AIAP — Governance (T1–T4) │
│ AIZP — Runtime Safety │
│ ★ AIKP — Knowledge Layer │ ← This protocol
│ AISOP — Execution │
│ SoulBot — Runtime │
└─────────────────────────────────────┘
HSAW — Axiom 0 (anchors all layers)
AIKP decides what to read; AISOP decides what to do; AIAP who is authorized; AIZP whether still aligned. Each independently upholds Axiom 0.
AIKP-Protocol/
├── specification/ # normative: AIKP_Protocol_cn.md, registry_cn.md, standards/
├── schemas/ # JSON Schemas (space/cache/focus/report)
├── examples/ # conformant example files
├── tests/ # validator + conformance suite
├── docs/ # English narrative
├── docs_cn/ # Chinese narrative: design intent / plan / lifecycle / related-work; plan/ = archived historical drafts (non-authoritative)
└── adrs/ # Architecture Decision Records
| Document | Description |
|---|---|
| Specification | Normative object model, schemas, validation, compliance |
| Registry | Canonical enum allocations (single source of truth) |
| Standards | External standards AIKP references / aligns with |
| Design Intent | Why AIKP exists; differentiation |
| Plan | Roadmap and decisions |
| Knowledge Lifecycle | Lifecycle / DB-style management analysis |
| Prior Art | Related work / prior-art positioning (cited) |
| CHANGELOG | Release history |
This protocol specification and any reference artifacts are provided for research and educational purposes only. They are experimental and not intended for production use. Use at your own risk. The authors assume no liability for any damages arising from use of this software. See LICENSE for full terms (Apache 2.0).
AIXP Labs aixp.dev
AIKP is the knowledge layer of the AIXP protocol family, alongside HSAW, AISOP, AIAP, AIZP, AIBP, AIVP, AIRP, AILP and the SoulBot reference runtime.
Align Axiom 0: Human Sovereignty and Wellbeing. AIKP v0.1.0. www.aikp.dev