MemForge Memory Layer

Example Use Cases

• 1

  Persistent, long-term memory across sessions.

• 2

  That require semantic and knowledge graph-based memory retrieval.

• 3

  Zero integration code and local embedding support.

Description

MemForge enables AI agents to have long-term conversational memory by storing, recalling, reflecting, and forgetting information through a single MCP server URL. It uses a single PostgreSQL 17 database with pgvector and Apache AGE for semantic search and knowledge graph capabilities. MemForge requires no SDK or integration code, supports local embeddings with no API keys, and achieves top performance on the LOCOMO benchmark.