# memory-lancedb MCP server

> Provides persistent long-term memory for AI agents using LanceDB with hybrid vector and keyword retrieval.

- Repository: https://github.com/cablate/memory-lancedb-mcp
- Page: https://hostedmcp.org/servers/cablate/memory-lancedb-mcp
- Transport: —
- Stars: 0
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `EMBEDDING_API_KEY` | no | no |  |
| `EMBEDDING_MODEL` | no | no |  |
| `MEMORY_LANCEDB_CONFIG` | no | no |  |

## Tools (20)

- `memory_recall`
- `memory_store`
- `memory_forget`
- `memory_update`
- `memory_merge`
- `memory_history`
- `memory_stats`
- `memory_list`
- `memory_lint`
- `self_improvement_log`
- `self_improvement_extract_skill`
- `self_improvement_review`
- `memory_visualize`
- `text`
- `vector`
- `category`
- `scope`
- `importance`
- `timestamp`
- `metadata`

## FAQ

### How does the memory retrieval work?

It uses a hybrid approach that searches by both semantic meaning and exact keywords simultaneously, then combines the results using RRF Fusion.

### Can I visualize my memory data?

Yes, you can use the memory_visualize tool to generate an interactive HTML knowledge graph that shows clusters and similarity edges.

### What happens if a new memory contradicts an old one?

The system automatically flags contradictions during the storage process to prevent silent overwrites.

### Does it support different embedding providers?

Yes, it works with any OpenAI-compatible embedding API, including OpenAI, Jina, Google Gemini, and local Ollama instances.