Memory MCP Client

Overview

Get current persona context from Memory MCP server using the get_context() tool. This skill provides a unified CLI client for all Memory MCP tools, enabling comprehensive interaction with the memory system.

When to Use

Always use at session start to:

•Retrieve current time and persona state
•Get memory statistics and recent activity
•Load conversation history for context continuity
•Understand persona's current situation before responding

Also use this skill when:

•Managing memories (create, search, update, delete)
•Working with items/equipment
•Checking memory statistics
•Debugging MCP integration

Available Tools

1. get_context

Retrieve current persona state, time, and memory statistics.

2. memory

Unified memory operations:

•create: Create new memory
•update: Update existing memory
•delete: Delete memory
•search: Search memories (semantic/keyword/hybrid/related/smart)
•stats: Get memory statistics
•check_routines: Detect recurring patterns
•anniversary: Manage anniversaries

3. item

Unified item operations:

•add: Add item to inventory
•remove: Remove item
•equip: Equip item
•unequip: Unequip item
•update: Update item details
•search: Search inventory
•history: Get equipment history
•memories: Get memories associated with item
•stats: Get item statistics

Quick Start

bash

# Script path: {skill_root}/scripts/memory_mcp_client.py
# Get context (always run at session start)
python <skill_root>/scripts/memory_mcp_client.py get_context

# Create a memory
python <skill_root>/scripts/memory_mcp_client.py memory create \
  --content "User likes strawberries" \
  --importance 0.8 \
  --emotion_type joy

# Search memories
python <skill_root>/scripts/memory_mcp_client.py memory search \
  --query "strawberries" \
  --mode semantic \
  --top_k 5

# Add and equip an item
python <skill_root>/scripts/memory_mcp_client.py item add \
  --name "Red Dress" \
  --category clothing \
  --description "Elegant red evening dress"

python <skill_root>/scripts/memory_mcp_client.py item equip --slot top --name "Red Dress"

# Equip multiple items at once
python <skill_root>/scripts/memory_mcp_client.py item equip \
  --equipment '{"top": "White Dress", "foot": "Sandals"}'

# Unequip a single slot
python <skill_root>/scripts/memory_mcp_client.py item unequip --slot top

# Unequip multiple slots
python <skill_root>/scripts/memory_mcp_client.py item unequip --slots top,foot

# Get item statistics
python <skill_root>/scripts/memory_mcp_client.py item stats

Note: Python client fully supports UTF-8 across all platforms (Windows/Linux/Mac) with automatic console encoding configuration.

Common Options

All commands support these options:

•--persona: Persona name (default: from config.json)
•--url: MCP server URL (default: from config.json)
•--format: Output format (text or json)

Configuration

Configuration is stored in references/config.json:

json

{
  "mcp_server": {
    "url": "http://localhost:26262",
    "timeout": 30,
    "retry_attempts": 3
  },
  "persona": {
    "default": "default",
    "description": "ペルソナ名はAuthorizationヘッダーで指定"
  },
  "paths": {
    "skill_root": "C:/Users/Owner/.claude/skills/memory-mcp/"
  },
  "settings": {
    "auto_get_context_on_session_start": true,
    "log_operations": true
  }
}

Configuration keys:

•mcp_server.url: Memory MCP server endpoint
•persona.default: Default persona name
•paths.skill_root: Absolute path to skill root directory (client script: {skill_root}/scripts/memory_mcp_client.py)
•settings.auto_get_context_on_session_start: Auto-call on session start

Memory Operations

Create Memory

bash

python <skill_root>/scripts/memory_mcp_client.py memory create \
  --content "Completed project milestone" \
  --importance 0.9 \
  --emotion_type accomplishment \
  --tags "work,achievement" \
  --context_tags "milestone" \
  --action_tag achievement

Update Memory

bash

python <skill_root>/scripts/memory_mcp_client.py memory update \
  --key "memory_20260209_123000" \
  --content "Updated content" \
  --importance 0.95

Search Memory

bash

# Semantic search (default)
python <skill_root>/scripts/memory_mcp_client.py memory search \
  --query "happy moments" \
  --mode semantic

# Keyword search
python <skill_root>/scripts/memory_mcp_client.py memory search \
  --query "Python" \
  --mode keyword

# Hybrid search (semantic 70% + keyword 30%)
python <skill_root>/scripts/memory_mcp_client.py memory search \
  --query "project" \
  --mode hybrid

# Time-filtered search
python <skill_root>/scripts/memory_mcp_client.py memory search \
  --query "achievements" \
  --date_range "昨日"

# Smart search (auto query expansion)
python <skill_root>/scripts/memory_mcp_client.py memory search \
  --query "いつものあれ" \
  --mode smart

Check Routines

bash

# Detect recurring patterns at current time
python <skill_root>/scripts/memory_mcp_client.py memory check_routines

Manage Anniversaries

bash

# List all anniversaries
python <skill_root>/scripts/memory_mcp_client.py memory anniversary

# Add anniversary
python <skill_root>/scripts/memory_mcp_client.py memory create \
  --content "First movie together" \
  --emotion_type joy \
  --importance 0.9 \
  --context_tags "first_time,anniversary"

# Delete anniversary
python <skill_root>/scripts/memory_mcp_client.py memory anniversary \
  --delete_key "memory_20260101_000000"

Item Operations

Add Items

bash

python <skill_root>/scripts/memory_mcp_client.py item add \
  --name "Blue Hat" \
  --category accessory \
  --description "Stylish blue hat" \
  --tags "casual,outdoor"

Equip/Unequip

bash

# Equip single item to a slot
python <skill_root>/scripts/memory_mcp_client.py item equip --slot accessory --name "Blue Hat"

# Equip multiple items at once
python <skill_root>/scripts/memory_mcp_client.py item equip \
  --equipment '{"top": "White Dress", "accessory": "Blue Hat", "foot": "Sandals"}'

# Unequip a single slot
python <skill_root>/scripts/memory_mcp_client.py item unequip --slot accessory

# Unequip multiple slots
python <skill_root>/scripts/memory_mcp_client.py item unequip --slots top,foot,accessory

Search Items

bash

# Search all items
python <skill_root>/scripts/memory_mcp_client.py item search

# Search by category
python <skill_root>/scripts/memory_mcp_client.py item search --category clothing

# Search equipped items only
python <skill_root>/scripts/memory_mcp_client.py item search --equipped True

Item History & Memories

bash

# Get equipment history
python <skill_root>/scripts/memory_mcp_client.py item history

# Get memories associated with an item
python <skill_root>/scripts/memory_mcp_client.py item memories --name "Red Dress"

Workflow

•Session Start → Run get_context to load current state
•Memory Management → Use memory operations for CRUD and search
•Item Management → Use item operations for inventory
•Monitoring → Check stats periodically with memory stats or item stats

REST API Endpoints

The CLI uses these MCP REST API endpoints:

•POST /mcp/v1/tools/get_context - Get context
•POST /mcp/v1/tools/memory - Memory operations
•POST /mcp/v1/tools/item - Item operations

All requests use Authorization: Bearer <persona> header.

Resources

scripts/memory_mcp_client.py

Unified CLI client for all MCP tools.

Features:

•All 3 MCP tools (get_context, memory, item)
•Comprehensive operation support
•JSON and text output formats
•Configuration file support
•Error handling and retries

references/config.json

Configuration file for MCP server connection and persona settings.

Customization:

•Update mcp_server.url for different server endpoints
•Change persona.default for different personas
•Adjust timeout and retry settings as needed