memory-knowledge-graph

name	memory-knowledge-graph
description	Knowledge graph-based persistent memory for AI agents — entities, relations, and semantic retrieval. Use when working with memory knowledge graph.
domain	mcp
author	oyi77
license	Apache-2.0
subdomain	mcp
tags	["ai-agent","graph","knowledge","mcp-server","memory","model-context-protocol","tool-integration"]
version	1.0.0

Overview

Persistent memory system using knowledge graphs for AI agents. Stores entities, relations, and observations as a graph, enabling semantic retrieval and reasoning over accumulated knowledge.

Capabilities

• Extract entities and relations from conversations and documents
• Store knowledge as a graph (nodes + edges + observations)
• Retrieve relevant knowledge via semantic search and graph traversal
• Consolidate and deduplicate knowledge over time
• Query knowledge with natural language

When to Use

Trigger phrases:

• "memory knowledge graph"

• "Knowledge graph-based persistent memory for AI agents — entities, relations, and"

• Building AI agents with long-term memory

• Need to remember user preferences, facts, and context across sessions

• Knowledge base that grows and improves over time

• Cross-referencing information across multiple sources

When NOT to Use

• Task is outside your authorization scope
• You need to implement controls (use implementing-* skills)
• Task is about analysis, not action (use analyzing-* skills)
• You don't have access to target systems
• Task requires compliance expertise (consult professionals)
• Task is about defense, not offense (use defensive skills)

Pseudo Code

# Example workflow for this skill
def execute(input_data):
    # Step 1: Validate input
    if not input_data:
        raise ValueError("Input data is required")

    # Step 2: Process core logic
    result = process(input_data)

    # Step 3: Validate output
    validate_output(result)

    return result

Entity Extraction

def extract_entities(text):
    prompt = f"""Extract entities and relations from:
{text}

Return JSON: {{"entities": [{{"name": "...", "type": "...", "observations": [...]}}], "relations": [{{"from": "...", "to": "...", "type": "..."}}]}}"""
    return llm.generate(prompt)

Knowledge Retrieval

def retrieve_relevant(query, graph, top_k=5):
    # 1. Semantic search on entity names and observations
    semantic_results = graph.semantic_search(query, top_k=top_k)
    
    # 2. Graph traversal - get connected entities
    related = []
    for entity in semantic_results:
        related.extend(graph.get_neighbors(entity.id, depth=2))
    
    # 3. Deduplicate and rank
    return rank_by_relevance(semantic_results + related, query)

Memory Consolidation

def consolidate(graph):
    # Merge duplicate entities
    duplicates = find_similar_entities(graph, threshold=0.9)
    for dup_group in duplicates:
        merge_entities(graph, dup_group)
    
    # Remove stale observations
    for entity in graph.entities:
        entity.observations = [o for o in entity.observations if o.relevance > 0.3]

Common Patterns

• Extract on every interaction: Build knowledge from every conversation
• Semantic + graph traversal: Combine vector search with graph walks for better retrieval
• Consolidate periodically: Merge duplicates and prune stale knowledge
• Entity types: Person, Organization, Concept, Event, Preference

How to Use

1. Invoke the skill when relevant domain keywords appear in the request
2. Provide required inputs as specified in the skill definition
3. Review the output for correctness before delivering to the user
4. Combine with related skills for complex multi-step workflows

Verification

After completing this skill, confirm:

• Output meets the defined quality and completeness requirements
• All prerequisites are verified and documented
• Error handling covers edge cases
• Results are accurate and actionable

Process

1. Analyze the task requirements
2. Apply domain expertise
3. Verify output quality

Anti-Rationalization Table

Rationalization	Reality
"I will just use curl"	MCP handles auth, retries, streaming, and type safety. Use the SDK.
"One mega-server is simpler"	Single-responsibility servers are easier to debug and maintain.
"MCP is just a wrapper"	MCP enables cross-platform tool sharing. It is infrastructure, not overhead.