Local Memory Mcp
by studiomeyer-iocommunityv2.3.1
8
4
An MCP server that stores persistent local memory in a single SQLite file on the user's machine and exposes tools for learning, entity observation, search, lifecycle queries, contradiction detection, and superseding stale facts. It uses local hybrid retrieval (FTS5/BM25 plus vector cosine) with multilingual embeddings and no API keys or cloud dependency.
From the registry: Persistent local AI memory. SQLite + FTS5 + vector search, knowledge graph. No cloud, no API keys.
Install & configure
Prerequisites
- requiredInstall Node.js and npxRequired to run the MCP server via the recommended npx install path.https://nodejs.org/
Install in your client
Paste into the AI you're chatting with — it will figure out the rest
Please install the `local-memory-mcp` MCP server into my current AI client (that's you).
Required prerequisites (do these first if not already done):
- **Install Node.js and npx** — Required to run the MCP server via the recommended npx install path. (https://nodejs.org/)
Canonical MCP server config (stdio transport):
- command: `npx`
- args: ["-y","@studiomeyer/local-memory-mcp"]
- optional environment variables:
- `MEMORY_DB_PATH`: Override the default SQLite database file location used to store local memory. (example: `<your-preferred-path>.sqlite`)
- `MEMORY_EMBED_DISABLED`: Disable embeddings and force FTS5-only search. (example: `1`)
- `MEMORY_EMBED_MODEL`: Override the local Transformers.js feature-extraction model used for embeddings. (example: `<your-model-name>`)
- `MEMORY_EMBED_CACHE_DIR`: Override the Transformers.js model cache directory. (example: `<your-cache-directory>`)
- `MEMORY_EMBED_DTYPE`: Set embedding model quantization type. (example: `q8`)
Note: Recommended portable install path is npx. No API keys are required. README also offers platform-specific .mcpb bundles for Claude Desktop, but the canonical MCP config is stdio via npx.
Add this MCP server to my current client's config in the correct format for you. If you need secrets or credentials I haven't provided, ASK me — do not invent values or leave raw placeholders. After adding it, tell me how to verify the server is connected.Note: Recommended portable install path is npx. No API keys are required. README also offers platform-specific .mcpb bundles for Claude Desktop, but the canonical MCP config is stdio via npx.
Tools (8)
memory_session_start
Starts a memory session so the server can load prior context.
memory_session_end
Ends a memory session.
memory_learn
Stores learnings, decisions, patterns, and mistakes.
memory_entity_observe
Records facts about people, projects, and tools as entity observations.
memory_search
Searches memory using keyword, vector, or hybrid retrieval.
memory_entity_open
Opens an entity with observations filtered by an as-of date or live view.
memory_contradictions
Finds semantically similar observations that conflict via negation or confidence drift.
memory_observation_supersede
Marks an older observation as superseded so it no longer appears in live views.
Environment variables
MEMORY_EMBED_DISABLEDForces FTS5-only search by disabling embeddings.MEMORY_EMBED_MODELOverrides the default Transformers.js embedding model.MEMORY_EMBED_CACHE_DIROverrides the cache directory used for embedding models.MEMORY_EMBED_DTYPESets embedding model quantization type such as fp32, fp16, q8, or q4.Packages
@studiomeyer/local-memory-mcp
npmv2.3.1transport: stdio
Details
- Language
- TypeScript
- License
- MIT
- Released
- Jun 23, 2026
- Last commit
- Jul 2, 2026
- Registry name
- io.github.studiomeyer-io/local-memory-mcp
Related Servers
Labelhead Artist Momentum
Trending hip-hop artist momentum scores across four cultural dimensions.
78.5k
PostHog MCP Server
Official PostHog MCP Server for product analytics, feature flags, experiments, and more.
37.7k
Openmetadata Mcp
Official OpenMetadata MCP: governed context and business semantics for AI assistants and agents.
14.9k