# GraphMemory-IDE MCP server

> Lets AI coding clients store and search long-term project memory in an on-device graph database with semantic vector search.

- Repository: https://github.com/elementalcollision/GraphMemory-IDE
- Page: https://hostedmcp.org/servers/elementalcollision/graphmemory-ide
- Transport: —
- Stars: 17
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `LADYBUG_DB_PATH` | no | no |  |

## FAQ

### What database does GraphMemory-IDE use for memory storage?

It runs on LadybugDB, an embedded columnar graph database accessed with Cypher, after migrating from KuzuDB.

### How does semantic search work in GraphMemory-IDE?

The /tools/topk endpoint embeds queries with sentence-transformers and runs top-K cosine similarity over node embedding columns.

### Which IDEs does GraphMemory-IDE support?

It ships plugins for VSCode, Cursor and Windsurf that connect to the MCP server.

### How do I start GraphMemory-IDE with Docker?

Clone the repo and run docker compose up -d inside the docker directory; the MCP server is then on port 8080.

### How is the GraphMemory-IDE server authenticated?

The FastAPI backend issues JWT tokens via an OAuth2 password flow at /auth/token, signed with EdDSA.