English | 中文

A graph-oriented interaction protocol designed specifically for Large Language Models,
bridging the gap between LLM and Knowledge Graph.

KIP (Knowledge Interaction Protocol) is a standard interaction protocol that bridges the gap between LLM (Probabilistic Reasoning Engine) and Knowledge Graph (Deterministic Knowledge Base). It is not merely a database interface, but a set of memory and cognitive operation primitives designed specifically for intelligent agents.

Large Language Models (LLMs) have demonstrated remarkable capabilities in general reasoning and generation. However, their "Stateless" essence results in a lack of long-term memory, while their probability-based generation mechanism often leads to uncontrollable "hallucinations" and knowledge obsolescence.

KIP was born to solve this problem through Neuro-Symbolic AI approach.

• 🧠 Memory Persistence: Transform conversations, observations, and reasoning results into structured "Knowledge Capsules"
• 📈 Knowledge Evolution: Complete CRUD and metadata management for autonomous learning and error correction
• 🔍 Explainable Interaction: Make every answer traceable and every decision logically transparent
• ⚡ LLM-Optimized: Protocol syntax optimized for Transformer architectures with native JSON structures

// Query: Find all drugs that treat headache
FIND(?drug.name)
WHERE {
  ?drug {type: "Drug"}
  (?drug, "treats", {name: "Headache"})
}
LIMIT 10

// Store: Create a new knowledge capsule
UPSERT {
  CONCEPT ?aspirin {
    {type: "Drug", name: "Aspirin"}
    SET ATTRIBUTES { molecular_formula: "C9H8O4", risk_level: 2 }
    SET PROPOSITIONS { ("treats", {type: "Symptom", name: "Headache"}) }
  }
}
WITH METADATA { source: "FDA", confidence: 0.95 }

// Explore: Discover schema
DESCRIBE PRIMER

A knowledge graph composed of Concept Nodes and Proposition Links, serving as the AI Agent's unified memory brain.

graph LR
    subgraph "Cognitive Nexus"
        A[Drug: Aspirin] -->|treats| B[Symptom: Headache]
        A -->|is_class_of| C[DrugClass: NSAID]
        A -->|has_side_effect| D[Symptom: Stomach Upset]
    end

KIP uses a self-describing schema where type definitions are stored within the graph itself:

• $ConceptType: Meta-type for defining concept node types
• $PropositionType: Meta-type for defining proposition predicates
• Domain: Organizational units for knowledge

This repository includes ready-to-use resources for building KIP-powered AI agents:

Pre-built knowledge capsules for bootstrapping your Cognitive Nexus:

kip-mcp-server - Model Context Protocol server that exposes KIP tools over stdio:

• Tools: execute_kip, list_logs
• Resources: kip://docs/SelfInstructions.md, kip://docs/KIPSyntax.md
• Prompt: kip_bootstrap for ready-to-inject system prompt

kip-cognitive-nexus - Publishable skill for AI agents:

• Python client script for anda_cognitive_nexus_server
• Complete syntax reference and agent workflow guide

A dedicated LLM layer that manages the Cognitive Nexus on behalf of business agents — no KIP knowledge required:

┌─────────────────────┐
│   Business Agent    │  ← No KIP knowledge needed
└────────┬────────────┘
         │ Natural Language
         ▼
┌─────────────────────┐
│    Hippocampus      │  ← Formation / Recall / Maintenance
└────────┬────────────┘
         │ KIP (KQL/KML/META)
         ▼
┌─────────────────────┐
│  Cognitive Nexus    │  ← Persistent Knowledge Graph
└─────────────────────┘

Full version history →

• 🔔 Products: Anda.AI | 天策 Celestian
• 💻 GitHub: LDC Labs
• 🏢 Company: Yiwen AI

Copyright © 2025 LDC Labs.

Licensed under the MIT License. See LICENSE for details.