Neo4j
Neo4j is the world's leading graph database, storing data as nodes, relationships, and properties rather than rows and columns. It uses the Cypher query language for expressing graph traversals, pattern matching, and aggregations. Neo4j exposes data via the Bolt binary protocol (used by all official drivers), an HTTP API for Cypher execution, and a REST Discovery API. It is available as Community Edition (open source), Enterprise Edition, and Neo4j Aura (fully managed cloud). A suite of four community-maintained MCP servers (from Neo4j Labs) enables AI agents to query, model, and manage graph data via natural language.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Username/password auth with optional SSO in Enterprise. Bolt protocol with TLS. Role-based access control in Enterprise edition. Aura (cloud) adds network isolation.
⚡ Reliability
Best When
Your data is fundamentally relational — fraud networks, knowledge graphs, recommendation graphs, org charts — where traversing relationships is the primary query pattern.
Avoid When
Your data is tabular with few meaningful relationships, or you need sub-millisecond OLAP aggregations over billions of rows.
Use Cases
- • Knowledge graph construction and traversal for AI reasoning and context enrichment
- • Fraud detection using multi-hop relationship traversal across transactional networks
- • Recommendation engines that use social graph topology (friends-of-friends, collaborative filtering)
- • Supply chain and network impact analysis using graph pathfinding algorithms (Dijkstra, PageRank)
- • Entity resolution and data lineage tracking across complex interconnected datasets
- • Semantic knowledge bases and ontologies powering RAG with structured relationship context
Not For
- • Relational/tabular data that doesn't have meaningful entity relationships
- • High-throughput OLAP analytics over billions of rows (use ClickHouse or BigQuery)
- • Simple key-value or document storage without graph traversal needs
- • Teams not willing to learn Cypher query language
Interface
Authentication
Native auth uses username/password via Bolt or HTTP Basic. Enterprise Edition adds LDAP, Kerberos, and SSO. Neo4j Aura uses username/password with connection URIs (neo4j+s:// scheme). Role-based access control (RBAC) with database-level and graph-element-level privileges available in Enterprise. Community Edition has limited role support.
Pricing
Community Edition is GPL v3 licensed — open source but single-instance only. Enterprise features (clustering, advanced security, RBAC) require paid license. Aura Free is permanent with modest limits suitable for development and small production workloads.
Agent Metadata
Known Gotchas
- ⚠ APOC plugin must be installed separately for schema introspection (apoc.meta.schema) — not available by default on Aura Free
- ⚠ MERGE creates new nodes if the pattern doesn't match exactly — incorrect property matching on MERGE leads to duplicate nodes silently
- ⚠ Cypher does not have a LIMIT clause on MATCH patterns (only on final RETURN) — missing LIMIT on write operations can affect entire graph
- ⚠ Graph traversal without depth limit ([:KNOWS*]) can cause runaway queries that exhaust memory — always specify max depth
- ⚠ HTTP API uses transaction endpoints — transactions must be explicitly committed or they auto-rollback after 60 seconds of inactivity
- ⚠ The mcp-neo4j suite is a Neo4j Labs experimental project — not production-supported; breaking changes possible
- ⚠ Natural language to Cypher translation by LLMs frequently gets relationship direction wrong — Neo4j is directional, wrong direction returns empty results
- ⚠ Community Edition only allows a single database per instance — multi-database support requires Enterprise license
Alternatives
Full Evaluation Report
Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Neo4j.
Scores are editorial opinions as of 2026-03-06.