
Memory Vault
Gives Claude four tools to store and retrieve memories from a self-hosted Postgres database with pgvector. It runs hybrid search combining semantic similarity and keyword matching, so you can recall past decisions or project context without re-explaining everything. The remember tool auto-classifies and embeds text, recall searches across memory spaces, forget soft-deletes by chunk ID, and memory_status reports database health. Runs on stdio transport and connects to your local Postgres instance. Useful if you want persistent context across Claude sessions without cloud storage or if you're building AI tools that need to remember state between conversations.
- GitHub stars
- 67
- tools
- 3
- auth
- Secret needed
- transport
- STDIO
- env vars
- 5
- last updated
- Oct 9, 2026
Tools
3 toolsPublic tool metadata for what this MCP can expose to an agent.
storeStore a JSON value under a key2 params
Store a JSON value under a key
keystringvalueobjectretrieveRetrieve a JSON value by key1 params
Retrieve a JSON value by key
keystringdeleteDelete a stored JSON value by key1 params
Delete a stored JSON value by key
keystringSponsors
Configuration
DB_HOST*Hostname of the Postgres+pgvector instance Memory Vault connects to (e.g. 'host.docker.internal' when Postgres runs on the host, 'localhost' when running natively, or the service name when both run in compose)
DB_PORT*default: 5432Postgres port
DB_NAME*default: memory_vaultPostgres database name
DB_USER*default: memory_vaultPostgres user with read/write access to the memory vault database
DB_PASSWORD*secretPostgres password for DB_USER
Readme
Memory Vault
The memory database for AI applications. Self-hosted Postgres + pgvector with hybrid search, MCP-native, and a knowledge graph baked in.
Details
- Repository
- github.com/mihaibuilds/memory-vault
- Website
- mihaibuilds.com
- Categories
- Databases, AI & LLM Tools, Search & Web Crawling
- Package
ghcr.io/mihaibuilds/memory-vault-mcp:1.6.1- Transport
- STDIO
- Auth
- Required
- Registry
- MCP registry, active