FAQ

General

What is MemNexus?

MemNexus is a universal memory layer for AI assistants. It stores memories in a graph database, makes them searchable via semantic search, and works with any AI platform through the CLI, SDK, or Model Context Protocol (MCP).

How is this different from ChatGPT Memory or Claude Projects?

Those are platform-specific. ChatGPT Memory only works in ChatGPT. Claude Projects only work in Claude. MemNexus works across all platforms — what Claude learns, Cursor can recall. It also provides structured knowledge graphs, pattern detection, and GraphRAG that built-in memories don't offer.

What platforms does MemNexus work with?

Any platform that supports MCP: Claude Desktop, Claude Code, Cursor, Windsurf, and more. For platforms without MCP support (like ChatGPT), you can use Custom GPT actions or the SDK.

Is MemNexus open source?

MemNexus is closed source. The CLI, SDK, and MCP server are published as npm packages.

Technical

What database does MemNexus use?

Neo4j, a graph database. This is a deliberate choice — graph databases excel at representing relationships between memories, topics, entities, and facts. Neo4j also provides vector indexes for semantic search.

What embedding model is used?

OpenAI's text-embedding-3-small (1536 dimensions). Embeddings are generated when memories are created and stored in Neo4j's vector index.

How does search work?

MemNexus runs a hybrid search combining:

Semantic search — Vector similarity using OpenAI embeddings
Keyword search — Full-text matching via Neo4j's Lucene indexes
Reciprocal Rank Fusion — Merges and re-ranks results from both

See Semantic Search for details.

What is GraphRAG?

Graph-based Retrieval Augmented Generation. Instead of just finding similar documents, GraphRAG traverses the knowledge graph to find connections, communities, and relationships that provide richer context for AI responses. See GraphRAG.

How many MCP tools does the server expose?

19 tools across 5 categories: Memory (6), Conversation (3), Topic (3), Pattern (4), and Advanced (3). See MCP Integration for the full reference.

Data & Privacy

Who can see my memories?

Only you. Each user's data is completely isolated. There is no mechanism to access another user's memories.

Is my data sent to third parties?

Memory content is sent to OpenAI to generate vector embeddings for search. OpenAI's API terms state that API inputs are not used for training. No other third-party services receive your data.

Can I export my data?

Yes. Use the CLI or API to export all your memories, facts, and conversations as JSON:

mx memories list --format json --limit 10000 > my-memories.json

Can I delete my data?

Yes. You can delete individual memories, facts, or conversations at any time. See Managing Memories.

Authentication

What format are API keys?

cmk_live_<id>.<secret>

The <id> identifies the key, and the <secret> is hashed server-side.

Where do I get an API key?

mx apikeys create --label "My Key"

Can I have multiple API keys?

Yes. Create separate keys for different environments or applications. Each key can be independently revoked.