Memforge Memory

Gives AI coding assistants a shared team engineering memory backed by PostgreSQL and pgvector semantic search.

What you can do

Memforge Memory provides a team engineering memory layer for AI coding assistants, exposing a large set of MCP tools (the description states 47) for storing and retrieving engineering knowledge. Retrieval uses pgvector semantic search, so entries are matched by meaning rather than keywords. It runs locally over stdio and connects to a PostgreSQL database for storage.

When to use it

Worth adding when a team wants coding assistants to recall shared engineering context, such as past decisions, notes, or project knowledge, across sessions and members. Useful if you already run PostgreSQL with pgvector and want semantic recall rather than a per-user local memory.

Requirements

Runs as a local process via the npm package @memforgeai/memory-service over stdio, on desktop only. Requires a PostgreSQL connection string in DATABASE_URL, an embedding API key in OPENAI_API_KEY, an OpenAI-compatible embedding base URL in OPENAI_BASE_URL, and an embedding model name in OPENAI_EMBEDDING_MODEL. Network access is needed for the embedding API.