The fastest way to get MemoryGraph working with Claude Code CLI.

Prerequisites

• Claude Code CLI installed (claude command available)
• Python 3.10 or higher
• pipx installed (recommended)

Quick Setup (3 Steps)

Step 1: Install MemoryGraph

# Install with pipx (handles PATH automatically)
pipx install memorygraphMCP

# Or with pip (ensure ~/.local/bin is in PATH)
pip install memorygraphMCP

Step 2: Add to Claude Code

# Add MCP server (user scope - available everywhere)
claude mcp add memorygraph memorygraph

# Or project scope (only in current directory)
claude mcp add memorygraph memorygraph --scope project

Step 3: Restart Claude Code

# Exit and restart Claude Code CLI
exit
claude

Verify Connection

# Check MCP servers status
claude mcp list

# You should see:
# memorygraph: connected

Important: Add Memory Directives

MemoryGraph won't automatically store memories. You need to add memory directives to tell Claude when and what to remember.

Add to your CLAUDE.md

Create or edit ~/.claude/CLAUDE.md (global) or .claude/CLAUDE.md (project):

## Memory Protocol

### REQUIRED: Before Starting Work
Use `recall_memories` before any task.

### REQUIRED: Automatic Storage
Store memories when:
- Fixing a bug → problem + solution
- Making architecture decisions
- Discovering reusable patterns
- Completing features

Always include tags, importance (0-1), and relationships.

→ Full Memory Protocol Guide with complete directives and examples.

Test Your First Memory

In Claude Code, try:

# Ask Claude to store a memory
"Store a memory that I successfully set up MemoryGraph"

# Recall it
"What do you remember about MemoryGraph setup?"

Optional: Install the Plugin

The MemoryGraph plugin adds auto-invoked skills, slash commands, and hooks for a seamless experience.

/plugin install memory-graph/memorygraph-plugin

With the plugin installed, you get:

• Auto-invoked skills - Solution capture when you fix errors, error tracking when you encounter issues
• Slash commands - /recall, /solve, /learn, /relate, /graph, /brief
• Smart hooks - Auto-load project context on session start, capture learnings before commits

→ Full Plugin Documentation

Configuration Options

User vs Project Scope

Custom Data Directory

# Use a custom database location
claude mcp add memorygraph memorygraph --env MEMORYGRAPH_DATA_DIR=/path/to/data

Enable Intelligence Features

# Install with intelligence extras
pipx install "memorygraphMCP[intelligence]"

# Then reconfigure
claude mcp remove memorygraph
claude mcp add memorygraph memorygraph

Cloud Backend (Multi-Device Sync)

Want your memories synced across all your devices? Use the MemoryGraph Cloud backend.

Step 1: Get Your API Key

Sign up at memorygraph.dev and create an API key from your dashboard.

Step 2: Configure Cloud Backend

# Remove existing local config (if any)
claude mcp remove memorygraph

# Add with cloud backend
claude mcp add --scope user memorygraph \
  --env MEMORYGRAPH_API_KEY=mg_your_key_here \
  -- memorygraph --backend cloud

Step 3: Restart Claude Code

exit
claude

Verify Cloud Connection

claude mcp list
# Should show: memorygraph: connected

Migrating Existing Local Memories

If you have existing local memories, export them first:

# Export from local SQLite
memorygraph export --format json --output my-memories.json

# Then configure cloud backend (above)
# Import to cloud
memorygraph import --format json --input my-memories.json

Troubleshooting

"Command not found: memorygraph"

# Ensure pipx PATH is set
pipx ensurepath
# Restart terminal, then try again

"Failed to connect" in MCP list

# Test memorygraph manually
memorygraph --health

# If it fails, check Python version
python3 --version  # Must be 3.10+

See the full Troubleshooting Guide for more solutions.