Pluribus introduces persistent memory service for AI agents to retain constraints and decisions across multiple sessions and users

Pluribus is a memory service using MCP protocol and HTTP with Postgres backend that stores structured memory including constraints, decisions, patterns, and failures

Solves the problem of agents losing context between runs since prompts and RAG don't preserve decisions, forcing re-derivation each time

Enables shared memory across multiple agents with tag-based recall and evaluation of proposed changes against existing memory

Supports local deployment or LAN operation, allowing agents to resume work with prior context and enforced constraints rather than ignored ones