
Memory MCP Server
A local memory layer that gives your assistant persistent, searchable context backed by SQLite. You get 13 tools covering CRUD, batch operations, FTS5 full-text search, and graph traversal with breadth-first recall across relationships. The standout feature is token-budget-aware retrieval: pass a token limit and it selects memories that fit, no manual pagination. Memories are SHA-256 addressed and support importance ranking, tags, and directed relationships. It also exposes resources for auto-completing memory hashes and a built-in usage guide. Runs over stdio with Node.js 24+, no HTTP endpoints, and includes Docker images. Reach for this when you need conversation context that persists across sessions and scales beyond simple key-value lookup.
- tools
- 13
- transport
- STDIO
- last updated
- Mar 1, 2026
Tools
13 toolsPublic tool metadata for what this MCP can expose to an agent.
store_memoryStore a new memory with content, tags, and optional type/importance. Idempotent — storing the same content+tags returns the existing hash with `created: false`.
Store a new memory with content, tags, and optional type/importance. Idempotent — storing the same content+tags returns the existing hash with `created: false`.
No parameter schema in public metadata yet.
store_memoriesStore multiple memories in one transaction (max 50 items).
Store multiple memories in one transaction (max 50 items).
No parameter schema in public metadata yet.
get_memoryRetrieve one memory by its SHA-256 hash.
Retrieve one memory by its SHA-256 hash.
No parameter schema in public metadata yet.
update_memoryUpdate content and optionally tags for an existing memory. Returns both hashes.
Update content and optionally tags for an existing memory. Returns both hashes.
No parameter schema in public metadata yet.
delete_memoryDelete one memory by hash. Cascades to related relationship rows.
Delete one memory by hash. Cascades to related relationship rows.
No parameter schema in public metadata yet.
delete_memoriesDelete multiple memories by hash in one transaction.
Delete multiple memories by hash in one transaction.
No parameter schema in public metadata yet.
search_memoriesFull-text search over memory content and tags using FTS5. Supports importance and type filters with cursor pagination.
Full-text search over memory content and tags using FTS5. Supports importance and type filters with cursor pagination.
No parameter schema in public metadata yet.
create_relationshipCreate a directed relationship edge between two memories. Idempotent.
Create a directed relationship edge between two memories. Idempotent.
No parameter schema in public metadata yet.
delete_relationshipDelete one directed relationship edge.
Delete one directed relationship edge.
No parameter schema in public metadata yet.
get_relationshipsRetrieve relationships for a memory, with optional direction filter.
Retrieve relationships for a memory, with optional direction filter.
No parameter schema in public metadata yet.
recallSearch memories by full-text query, then traverse the relationship graph up to `depth` hops via BFS. Emits MCP progress notifications per hop.
Search memories by full-text query, then traverse the relationship graph up to `depth` hops via BFS. Emits MCP progress notifications per hop.
No parameter schema in public metadata yet.
retrieve_contextSearch memories and return relevance-ranked results that fit within a caller-specified token budget.
Search memories and return relevance-ranked results that fit within a caller-specified token budget.
No parameter schema in public metadata yet.
memory_statsReturn aggregate memory and relationship stats. Takes no input.
Return aggregate memory and relationship stats. Takes no input.
No parameter schema in public metadata yet.
Sponsors
Readme
Details
- Repository
- github.com/j0hanz/memory-mcp
- Website
- github.com/j0hanz/memory-mcp#readme
- Categories
- Databases, AI & LLM Tools
- Package
@j0hanz/memory-mcp- Transport
- STDIO
- Registry
- MCP registry, active