Verified

Memory Service MCP Server

Persistent shared memory backend for AI agents using knowledge graphs and REST APIs.

Data last scanned today · Reviewed today

651 tool★ 1,927PythonApache-2.0

Overview

This Python-based service provides a persistent memory layer for AI agents, allowing them to store and retrieve project context, architecture details, and causal knowledge graphs. It eliminates the need for repeated setup by maintaining long-term memory across chat sessions for various agent frameworks including LangGraph, CrewAI, and AutoGen. The system operates via a self-hosted architecture that includes a REST API, dashboard, and OAuth 2.0 support. By treating memories as nodes in an interactive 3D knowledge graph, it allows agents to recall information in 5ms without incurring cloud lock-in or additional API usage costs.

Our verdict

With a security score of 50/100, this server is classified as high risk. The rating is primarily due to the fact that the authentication method and read-only mode have not yet been editorially reviewed. While it benefits from zero direct dependencies and a large contributor base, users should exercise caution when implementing this in environments where data access control and identity verification are critical until a formal audit is completed.

Zero direct dependencies reduce supply-chain surface area
Active development with commits as recent as two days ago
Broad compatibility with major frameworks like LangGraph and CrewAI
Open-source Apache-2.0 licensing

Security-sensitive features like authentication remain unreviewed
Lack of official vendor maintenance increases oversight burden
Read-only operational modes have not been verified

Tools

Tool	Description	Risk
get_memory_subgraph	Retrieve the subgraph for a specific memory.	low

Compatibility

Client	Local	Docker	Remote	Read-only
Claude Desktop
Cursor
VS Code
Windsurf
ChatGPT

Frequently asked questions

›Can I use this memory service without an MCP client?

Yes, the service includes a framework-agnostic REST API that allows it to function with any HTTP client, including LangGraph, CrewAI, and AutoGen.

›Does this service send my memory data to the cloud?

No, the service uses local ONNX embeddings, ensuring that your memory data never leaves your local infrastructure.

›How do I scope memory retrieval to specific agents?

You can auto-tag memories by providing an X-Agent-ID header, which allows you to filter search results to a specific agent's history.

›Does the server require authentication for API access?

The server supports OAuth 2.0 with DCR, though you can enable anonymous access for local development by setting the MCP_ALLOW_ANONYMOUS_ACCESS environment variable.

›How can I monitor storage or deletion activity in real-time?

The server provides SSE events that deliver real-time notifications whenever any agent stores or deletes a memory.