Neo4j Memory MCP Server

Neo4j Memory MCP server (part of the neo4j-contrib MCP suite) providing AI agents with graph-based persistent memory — storing entities and their relationships as a knowledge graph in Neo4j, enabling semantic memory retrieval, managing entity attributes and connections, and integrating Neo4j's graph database as a long-term memory backend for AI agent knowledge graphs.

Evaluated Mar 06, 2026 (0d ago) vcurrent
Homepage ↗ Repo ↗ Databases neo4j memory knowledge-graph mcp-server graph-database persistent-memory entities
⚙ Agent Friendliness
79
/ 100
Can an agent use this?
🔒 Security
81
/ 100
Is it safe for agents?
⚡ Reliability
77
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
78
Documentation
80
Error Messages
75
Auth Simplicity
82
Rate Limits
80

🔒 Security

TLS Enforcement
92
Auth Strength
80
Scope Granularity
72
Dep. Hygiene
80
Secret Handling
82

Bolt TLS. User/pass. SOC2/GDPR. Neo4j-contrib. Store creds in env var. Sensitive memory data.

⚡ Reliability

Uptime/SLA
80
Version Stability
78
Breaking Changes
75
Error Recovery
75
AF Security Reliability

Best When

An agent needs structured relationship-aware memory — tracking how entities connect to each other rather than just storing isolated facts, using Neo4j's graph model for rich knowledge representation.

Avoid When

You need simple persistent memory without relationship tracking — use lighter alternatives. Or if you prefer vector-based semantic search.

Use Cases

  • Storing entities and relationships as persistent knowledge graph memory
  • Building and querying agent knowledge graphs from personal assistant agents
  • Maintaining long-term user preference graphs from personalization agents
  • Tracking entity relationships over time from analysis agents
  • Implementing semantic memory with graph traversal for reasoning agents
  • Building organizational knowledge graphs from enterprise knowledge agents

Not For

  • Teams without Neo4j instance (AuraDB free tier available)
  • Simple key-value memory (use lighter alternatives like mcp-memory-keeper)
  • Vector similarity search (use ChromaDB or LanceDB for embedding-based memory)

Interface

REST API
Yes
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
Yes
Webhooks
No

Authentication

Methods: username_password
OAuth: No Scopes: No

Neo4j username/password required. Uses Neo4j Bolt protocol. AuraDB free tier supported. Same auth as standard mcp-neo4j-server.

Pricing

Model: freemium
Free tier: Yes
Requires CC: No

AuraDB Free is sufficient for agent memory use cases. Community/Enterprise self-hosted is free. MCP from neo4j-contrib.

Agent Metadata

Pagination
none
Idempotent
Partial
Retry Guidance
Not documented

Known Gotchas

  • Entity naming and relationship types must be consistent — agents should establish naming conventions
  • Graph can grow unbounded — implement cleanup or archival strategies for long-running agents
  • Cypher queries required for complex memory retrieval — simpler than raw Neo4j MCP
  • Part of neo4j-contrib MCP suite — same high quality as mcp-neo4j-server
  • Memory retrieval is relationship-traversal based — different paradigm from vector similarity
  • AuraDB Free tier node limit (50K) is sufficient for most agent memory use cases

Alternatives

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Neo4j Memory MCP Server.

$99

Scores are editorial opinions as of 2026-03-06.

5220
Packages Evaluated
26151
Need Evaluation
173
Need Re-evaluation
Community Powered