# Sunbeam Memory MCP server

> Provides a personal semantic memory server that allows AI assistants to store and search facts, code, and documents using vector embeddings.

- Repository: https://github.com/sunbeamdotpt/memory
- Page: https://hostedmcp.org/servers/sunbeamdotpt/memory
- Transport: —
- Stars: 23
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `MCP_AUTH_TOKEN` | no | no |  |
| `MCP_MEMORY_BASE_DIR` | no | no |  |
| `MCP_OIDC_AUDIENCE` | no | no |  |
| `MCP_OIDC_ISSUER` | no | no |  |

## FAQ

### How does the server handle file updates?

The server watches directories and automatically ingests files like code, PDFs, and text. For git repositories, it tracks state so edits on the same branch update existing facts in place.

### Can I access my memory from different machines?

Yes, you can run the server in HTTP mode on a VPS or home server to access your memory from any machine or AI client.

### How do I back up my stored data?

Since the data is stored in a self-contained SQLite database, you can back it up simply by copying the semantic.db file.

### Does it support authentication for remote use?

Yes, you can use a simple bearer token via MCP_AUTH_TOKEN or integrate with an OIDC provider like Keycloak or Auth0.