MemoryGraph Cloud provides multi-device sync, automatic backups, and zero maintenance. Access your memories from any device with your coding agent.

Benefits

• Multi-device sync - Access memories from laptop, desktop, or cloud IDE
• Zero maintenance - No database setup or backups needed
• Team collaboration - Share knowledge graphs with your team (coming soon)
• Automatic backups - Your data is always safe
• Native graph queries - Full Cypher support via FalkorDB
• Graph intelligence - MCR² optimization, similarity detection, transitive inference
• Memory consolidation - Automatic organization inspired by biological memory

Cloud Intelligence Features

The cloud backend includes advanced graph intelligence features that automatically optimize your memory graph structure.

MCR² (Maximal Coding Rate Reduction)

MCR² is an information-theoretic metric that measures how well your memories are organized. Higher rate reduction means memories are better clustered by semantic type.

• Total Rate - Entropy of all memories as a single distribution
• Class Rate - Sum of entropies within each memory type cluster
• Rate Reduction - The difference: lower is better organized

The cloud backend tracks MCR² scores and uses them to guide consolidation decisions. Prometheus metrics available: mcr2_rate_reduction, mcr2_class_count.

Automatic Similarity Detection

When you store a memory, the cloud backend automatically finds similar memories and creates relationships:

• 0.70+ cosine similarity - Creates a SIMILAR_TO relationship
• 0.90+ cosine similarity - Flagged as merge candidate for consolidation
• Top-10 comparison - Checks against the 10 most similar memories
• Async processing - Runs in batches, doesn't impact write latency

Transitive Inference

The cloud backend automatically infers new relationships from existing ones. If A relates to B and B relates to C, the system can infer A relates to C.

Inference Rules

• CAUSES chains - If A CAUSES B and B CAUSES C, infer A LEADS_TO C
• SOLVES chains - If solution SOLVES problem and problem CAUSES error, link solution to error
• SIMILAR_TO transitivity - Similar memories form clusters

Confidence Degradation

Inferred relationships include a confidence score that decreases with path depth:

# Confidence calculation
confidence = 1 / depth

# Examples:
depth 1 (direct): confidence = 1.0
depth 2 (one hop): confidence = 0.5
depth 3 (two hops): confidence = 0.33

Transitive inference runs asynchronously in batches and doesn't impact write latency.

Memory Consolidation

Inspired by biological memory consolidation during sleep, the cloud backend periodically reorganizes memories to improve organization and reduce redundancy.

Consolidation Operations

Biological Inspiration

• Sharp-wave ripples - Quick replay and connection (similarity detection)
• REM consolidation - Pattern extraction and abstraction
• Synaptic homeostasis - Pruning weak connections
• Sleep spindles - Strengthening important memories

Consolidation runs on a configurable schedule (default: daily) and tracks MCR² scores before and after to measure effectiveness.

Quick Setup

Step 1: Get Your API Key

1. Sign up at memorygraph.dev
2. Go to your Dashboard
3. Click "Create API Key"
4. Copy the key (starts with mg_)

Step 2: Configure Your Client

Claude Code CLI

# 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

# Restart Claude Code
exit
claude

Claude Desktop / Cursor / Other MCP Clients

Add to your MCP config file:

{
  "mcpServers": {
    "memorygraph": {
      "command": "memorygraph",
      "args": ["--backend", "cloud"],
      "env": {
        "MEMORYGRAPH_API_KEY": "mg_your_key_here"
      }
    }
  }
}

Environment Variables

export MEMORYGRAPH_API_KEY=mg_your_key_here
memorygraph --backend cloud

Step 3: Verify Connection

# Check health
memorygraph --health

# In Claude Code
claude mcp list
# Should show: memorygraph: connected

Migrating from Local to Cloud

If you have existing memories in a local SQLite database, you can migrate them to the cloud.

Option A: Export/Import (Recommended)

# 1. Export from local (while still using local backend)
memorygraph export --format json --output my-memories.json

# 2. Configure cloud backend (see above)

# 3. Import to cloud
memorygraph import --format json --input my-memories.json

Option B: Built-in Migration

# Migrate directly from SQLite to cloud
memorygraph migrate \
  --from sqlite \
  --to cloud \
  --dry-run  # Remove to execute

# With verification
memorygraph migrate \
  --from sqlite \
  --to cloud \
  --verbose

Configuration Options

Offline Mode

The cloud backend includes a circuit breaker for network resilience. If the connection is temporarily unavailable, operations will retry automatically with exponential backoff.

Data Privacy

• Your memories are isolated in your own graph namespace
• Data is encrypted in transit (TLS) and at rest
• API keys can be revoked at any time from your dashboard
• Export your data anytime with memorygraph export

Troubleshooting

"Unable to connect to cloud backend"

# Check API key is set
echo $MEMORYGRAPH_API_KEY

# Test connection
curl -H "X-API-Key: $MEMORYGRAPH_API_KEY" \
  https://graph-api.memorygraph.dev/health

"Authentication failed"

1. Verify your API key is correct (starts with mg_)
2. Check the key hasn't been revoked in your dashboard
3. Ensure there are no extra spaces or newlines in the key

"Rate limit exceeded"

Free tier includes 1000 requests/day. Upgrade your plan for higher limits. See Pricing for details.

Switching Back to Local

You can switch back to local storage at any time:

# Export from cloud first
memorygraph export --format json --output cloud-backup.json

# Reconfigure for local
claude mcp remove memorygraph
claude mcp add memorygraph memorygraph

# Import to local
memorygraph import --format json --input cloud-backup.json