🧠🕸️ AgensGraph Knowledge Graph Memory MCP Server
🌟 Overview
A Model Context Protocol (MCP) server implementation that provides persistent memory capabilities through AgensGraph graph database integration.
By storing information in a graph structure, this server maintains complex relationships between entities as memory nodes and enables long-term retention of knowledge that can be queried and analyzed across multiple conversations or sessions.
The MCP server leverages AgensGraph's graph database capabilities to create an interconnected knowledge base that serves as an external memory system. Through Cypher queries, it allows exploration and retrieval of stored information, relationship analysis between different data points, and generation of insights from the accumulated knowledge.
🕸️ Graph Schema
Memory- A node representing an entity with a name, type, and observations.Relationship- A relationship between two entities with a type.
📦 Components
🔧 Tools
The server offers these core tools:
🔎 Query Tools
-
read_graph- Read the knowledge graph
- Input:
limit(int, optional): max entities to return (default 1000). Relations reference entities by name, so the result stays coherent when capped.
- Returns:
{ "entities": [...], "relations": [...], "truncated": bool }. Whentruncatedis true the memory has more — narrow withsearch_memoriesor raise the limit. (Default cap configurable viaAGENSGRAPH_MEMORY_LIMIT.)
-
search_memories- Search for nodes based on a query
- Input:
query(string): Search query matching names, types, observationslimit(int, optional): max matching entities to return (default 1000)
- Returns: matching subgraph as
{ "entities", "relations", "truncated" }
-
find_memories_by_name- Find specific nodes by name
- Input:
names(array of strings): Entity names to retrieve
- Returns: Subgraph with specified nodes
♟️ Entity Management Tools
-
create_entities- Create multiple new entities in the knowledge graph
- Input:
entities: Array of objects with:name(string): Name of the entitytype(string): Type of the entityobservations(array of strings): Initial observations about the entity
- Returns: Created entities
-
delete_entities- Delete multiple entities and their associated relations
- Input:
entityNames(array of strings): Names of entities to delete
- Returns: Success confirmation
🔗 Relation Management Tools
-
create_relations- Create multiple new relations between entities
- Input:
relations: Array of objects with:source(string): Name of source entitytarget(string): Name of target entityrelationType(string): Type of relation
- Returns: Created relations
-
delete_relations- Delete multiple relations from the graph
- Input:
relations: Array of objects with same schema as create_relations
- Returns: Success confirmation
📝 Observation Management Tools
-
add_observations- Add new observations to existing entities
- Input:
observations: Array of objects with:entityName(string): Entity to add tocontents(array of strings): Observations to add
- Returns: Added observation details
-
delete_observations- Delete specific observations from entities
- Input:
deletions: Array of objects with:entityName(string): Entity to delete fromobservations(array of strings): Observations to remove
- Returns: Success confirmation
🔧 Usage with Claude Desktop
💾 Installation
pip install mcp-agensgraph-memory
⚙️ Configuration
Add the server to your claude_desktop_config.json:
"mcpServers": {
"agensgraph": {
"command": "uvx",
"args": [
"mcp-agensgraph-memory",
"--db-url",
"postgresql://localhost:5432",
"--username",
"<your-username>",
"--password",
"<your-password>",
"--database",
"<your-database>",
"--graphname",
"memory"
]
}
}
Alternatively, you can set environment variables:
"mcpServers": {
"agensgraph": {
"command": "uvx",
"args": [ "mcp-agensgraph-memory" ],
"env": {
"AGENSGRAPH_URL": "postgresql://localhost:5432",
"AGENSGRAPH_USERNAME": "<your-username>",
"AGENSGRAPH_PASSWORD": "<your-password>",
"AGENSGRAPH_DATABASE": "<your-database>",
"AGENSGRAPH_GRAPH_NAME": "memory"
}
}
}
Namespacing
For multi-tenant deployments, add --namespace to prefix tool names:
"args": [ "mcp-agensgraph-memory", "--namespace", "myapp", "--db-url", "..." ]
Tools become: myapp-read_graph, myapp-create_entities, etc.
Can also use AGENSGRAPH_NAMESPACE environment variable.
🌐 HTTP Transport Mode
The server supports HTTP transport for web-based deployments and microservices:
# Basic HTTP mode (defaults: host=127.0.0.1, port=8000, path=/mcp/)
mcp-agensgraph-memory --transport http
# Custom HTTP configuration
mcp-agensgraph-memory --transport http --server-host 127.0.0.1 --server-port 8080 --server-path /api/mcp/
Environment variables for HTTP configuration:
export AGENSGRAPH_TRANSPORT=http
export AGENSGRAPH_MCP_SERVER_HOST=127.0.0.1
export AGENSGRAPH_MCP_SERVER_PORT=8080
export AGENSGRAPH_MCP_SERVER_PATH=/api/mcp/
export AGENSGRAPH_NAMESPACE=myapp
mcp-agensgraph-memory
🔄 Transport Modes
The server supports three transport modes:
- STDIO (default): Standard input/output for local tools and Claude Desktop
- SSE: Server-Sent Events for web-based deployments
- HTTP: Streamable HTTP for modern web deployments and microservices
🔒 Security Protection
The server includes comprehensive security protection with secure defaults that protect against common web-based attacks while preserving full MCP functionality when using HTTP transport.
🛡️ DNS Rebinding Protection
TrustedHost Middleware validates Host headers to prevent DNS rebinding attacks:
Secure by Default:
- Only
localhostand127.0.0.1hosts are allowed by default
Environment Variable:
export AGENSGRAPH_MCP_SERVER_ALLOWED_HOSTS="example.com,www.example.com"
🌐 CORS Protection
Cross-Origin Resource Sharing (CORS) protection blocks browser-based requests by default:
Environment Variable:
export AGENSGRAPH_MCP_SERVER_ALLOW_ORIGINS="https://example.com,https://app.example.com"
🔧 Complete Security Configuration
Development Setup:
mcp-agensgraph-memory --transport http \
--allowed-hosts "localhost,127.0.0.1" \
--allow-origins "http://localhost:3000"
Production Setup:
mcp-agensgraph-memory --transport http \
--allowed-hosts "example.com,www.example.com" \
--allow-origins "https://example.com,https://app.example.com"
🚀 Development
📦 Prerequisites
- Install
uv(Universal Virtualenv):
# Using pip
pip install uv
# Using Homebrew on macOS
brew install uv
# Using cargo (Rust package manager)
cargo install uv
- Clone the repository and set up development environment:
# Create and activate virtual environment using uv
uv venv
source .venv/bin/activate # On Unix/macOS
.venv\Scripts\activate # On Windows
# Install dependencies including dev dependencies
uv sync
🔧 Environment Variables
| Variable | Default | Description |
|---|---|---|
AGENSGRAPH_URL |
postgresql://localhost:5432 |
AgensGraph connection URL (host:port only) |
AGENSGRAPH_USERNAME |
agens |
AgensGraph username |
AGENSGRAPH_PASSWORD |
agens |
AgensGraph password |
AGENSGRAPH_DB |
agens |
AgensGraph database name |
AGENSGRAPH_GRAPH_NAME |
memory |
AgensGraph graph name |
AGENSGRAPH_TRANSPORT |
stdio (local), http (remote) |
Transport protocol (stdio, http, or sse) |
AGENSGRAPH_MCP_SERVER_HOST |
127.0.0.1 (local) |
Host to bind to |
AGENSGRAPH_MCP_SERVER_PORT |
8000 |
Port for HTTP/SSE transport |
AGENSGRAPH_MCP_SERVER_PATH |
/mcp/ |
Path for accessing MCP server |
AGENSGRAPH_MCP_SERVER_ALLOW_ORIGINS |
(empty - secure by default) | Comma-separated list of allowed CORS origins |
AGENSGRAPH_MCP_SERVER_ALLOWED_HOSTS |
localhost,127.0.0.1 |
Comma-separated list of allowed hosts (DNS rebinding protection) |
AGENSGRAPH_NAMESPACE |
(empty - no prefix) | Namespace prefix for tool names (e.g., myapp-read_graph) |
📄 License
This MCP server is licensed under the MIT License. This means you are free to use, modify, and distribute the software, subject to the terms and conditions of the MIT License. For more details, please see the LICENSE file in the project repository.
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file mcp_agensgraph_memory-0.2.0.tar.gz.
File metadata
- Download URL: mcp_agensgraph_memory-0.2.0.tar.gz
- Upload date:
- Size: 26.3 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/7.0.0 CPython/3.13.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
5c9d0bc09e4471ba9176924b0532a5ef7b296b5b4b82e21c85b2231a35ba441c
|
|
| MD5 |
14415ea1723ee5b141ec5fc6194e9615
|
|
| BLAKE2b-256 |
b4a32a82ff8b4b757332ed8dc7c9f0a094bf972554d88b543794244feb7f85b9
|
File details
Details for the file mcp_agensgraph_memory-0.2.0-py3-none-any.whl.
File metadata
- Download URL: mcp_agensgraph_memory-0.2.0-py3-none-any.whl
- Upload date:
- Size: 20.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/7.0.0 CPython/3.13.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
832684f12ca845960ba010ed2896ff7875fa99268e3bf0f5bc372845f402251b
|
|
| MD5 |
a8e802e5e2d8a7242e0fea3837fe1925
|
|
| BLAKE2b-256 |
64bb6d71babed158b62d7aa7dc7f6d4446af2eed63542a4ead8ee162f6b0de7d
|