mcp/neo4j-memory repository overview

⁠Neo4j Memory MCP Server

Provide persistent memory capabilities through Neo4j graph database integration.

What is an MCP Server?⁠

⁠MCP Info

⁠Image Building Info

Attribute	Details
Dockerfile	https://github.com/neo4j-contrib/mcp-neo4j/blob/1d3c88743e2e44e374639e79aa71aac90c6012e1/servers/mcp-neo4j-memory/Dockerfile⁠
Commit	1d3c88743e2e44e374639e79aa71aac90c6012e1
Docker Image built by	Docker Inc.
Docker Scout Health Score
Verify Signature	COSIGN_REPOSITORY=mcp/signatures cosign verify mcp/neo4j-memory --key https://raw.githubusercontent.com/docker/keyring/refs/heads/main/public/mcp/latest.pub
Licence	MIT License

⁠Available Tools (9)

⁠Tools Details

⁠Tool: add_observations

Add new observations/facts to existing entities in the knowledge graph.

Appends new observations to entities that already exist. The entity must be present in the graph before adding observations. Each observation should be a distinct fact.

Returns: list[dict]: Details about the added observations including entity name and new facts

Example call: { "observations": [ { "entityName": "Alice Johnson", "observations": ["Promoted to Senior Engineer", "Completed AWS certification"] }, { "entityName": "Microsoft", "observations": ["Launched new AI products", "Stock price increased 15%"] } ] }

This tool interacts with external entities.

⁠Tool: create_entities

Create multiple new entities in the knowledge graph.

Creates new memory entities with their associated observations. If an entity with the same name already exists, this operation will merge the observations with existing ones.

Returns: list[Entity]: The created entities with their final state

Example call: { "entities": [ { "name": "Alice Johnson", "type": "person", "observations": ["Software engineer", "Lives in Seattle", "Enjoys hiking"] }, { "name": "Microsoft", "type": "company", "observations": ["Technology company", "Headquartered in Redmond, WA"] } ] }

This tool interacts with external entities.

⁠Tool: create_relations

Create multiple new relationships between existing entities in the knowledge graph.

Creates directed relationships between entities that already exist. Both source and target entities must already be present in the graph. Use descriptive relationship types.

Returns: list[Relation]: The created relationships

Example call: { "relations": [ { "source": "Alice Johnson", "target": "Microsoft", "relationType": "WORKS_AT" }, { "source": "Alice Johnson", "target": "Seattle", "relationType": "LIVES_IN" } ] }

This tool interacts with external entities.

⁠Tool: delete_entities

Delete entities and all their associated relationships from the knowledge graph.

Permanently removes entities from the graph along with all relationships they participate in. This is a destructive operation that cannot be undone. Entity names must match exactly.

Returns: str: Success confirmation message

Example call: { "entityNames": ["Old Company", "Outdated Person"] }

Warning: This will delete the entities and ALL relationships they're involved in.

This tool may perform destructive updates.

This tool is idempotent. Repeated calls with same args have no additional effect.

This tool interacts with external entities.

⁠Tool: delete_observations

Delete specific observations from existing entities in the knowledge graph.

Removes specific observation texts from entities. The observation text must match exactly what is stored. The entity will remain but the specified observations will be deleted.

Returns: str: Success confirmation message

Example call: { "deletions": [ { "entityName": "Alice Johnson", "observations": ["Old job title", "Outdated phone number"] }, { "entityName": "Microsoft", "observations": ["Former CEO information"] } ] }

Note: Observation text must match exactly (case-sensitive) to be deleted.

This tool may perform destructive updates.

This tool is idempotent. Repeated calls with same args have no additional effect.

This tool interacts with external entities.

⁠Tool: delete_relations

Delete specific relationships between entities in the knowledge graph.

Removes relationships while keeping the entities themselves. The source, target, and relationship type must match exactly for deletion. This only affects the relationships, not the entities they connect.

Returns: str: Success confirmation message

Example call: { "relations": [ { "source": "Alice Johnson", "target": "Old Company", "relationType": "WORKS_AT" }, { "source": "John Smith", "target": "Former City", "relationType": "LIVES_IN" } ] }

Note: All fields (source, target, relationType) must match exactly for deletion.

This tool may perform destructive updates.

This tool is idempotent. Repeated calls with same args have no additional effect.

This tool interacts with external entities.

⁠Tool: find_memories_by_name

Find specific entities by their exact names.

Retrieves entities that exactly match the provided names, along with all their relationships and connected entities. Use this when you know the exact entity names.

Returns: KnowledgeGraph: Subgraph containing the specified entities and their relationships

Example call: { "names": ["Alice Johnson", "Microsoft", "Seattle"] }

This retrieves the entities with exactly those names plus their connections.

This tool is read-only. It does not modify its environment.

This tool interacts with external entities.

⁠Tool: read_graph

Read the entire knowledge graph with all entities and relationships.

Returns the complete memory graph including all stored entities and their relationships. Use this to get a full overview of stored knowledge.

Returns: KnowledgeGraph: Complete graph with all entities and relations

Example response: { "entities": [ {"name": "John Smith", "type": "person", "observations": ["Works at Neo4j"]}, {"name": "Neo4j Inc", "type": "company", "observations": ["Graph database company"]} ], "relations": [ {"source": "John Smith", "target": "Neo4j Inc", "relationType": "WORKS_AT"} ] }

⁠Tool: search_memories

Search for entities in the knowledge graph using fulltext search.

Searches across entity names, types, and observations using Neo4j's fulltext index. Returns matching entities and their related connections. Supports partial matches and multiple search terms.

Returns: KnowledgeGraph: Subgraph containing matching entities and their relationships

Example call: { "query": "engineer software" }

This searches for entities containing "engineer" or "software" in their name, type, or observations.

This tool is read-only. It does not modify its environment.

This tool interacts with external entities.

⁠Use this MCP Server

{
  "mcpServers": {
    "neo4j-memory": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "NEO4J_URL",
        "-e",
        "NEO4J_USERNAME",
        "-e",
        "NEO4J_DATABASE",
        "-e",
        "NEO4J_TRANSPORT",
        "-e",
        "NEO4J_MCP_SERVER_HOST",
        "-e",
        "NEO4J_MCP_SERVER_PORT",
        "-e",
        "NEO4J_MCP_SERVER_PATH",
        "-e",
        "NEO4J_MCP_SERVER_ALLOW_ORIGINS",
        "-e",
        "NEO4J_MCP_SERVER_ALLOWED_HOSTS",
        "-e",
        "NEO4J_PASSWORD",
        "mcp/neo4j-memory"
      ],
      "env": {
        "NEO4J_URL": "bolt://host.docker.internal:7687",
        "NEO4J_USERNAME": "neo4j",
        "NEO4J_DATABASE": "neo4j",
        "NEO4J_TRANSPORT": "http",
        "NEO4J_MCP_SERVER_HOST": "0.0.0.0",
        "NEO4J_MCP_SERVER_PORT": "8000",
        "NEO4J_MCP_SERVER_PATH": "/mcp/",
        "NEO4J_MCP_SERVER_ALLOW_ORIGINS": "example.com,www.example.com",
        "NEO4J_MCP_SERVER_ALLOWED_HOSTS": "localhost,127.0.0.1",
        "NEO4J_PASSWORD": "password"
      }
    }
  }
}

Why is it safer to run MCP Servers with Docker?⁠