OB1: a unified persistent memory layer and open protocol for sharing context across multiple AI tools

What it solves

Open Brain (OB1) provides a unified, persistent memory layer for all the AI tools you use. Instead of each AI (like Claude, ChatGPT, or Cursor) having its own isolated memory or context, OB1 creates a single database that acts as a shared "brain" across different AI gateways and chat channels, eliminating the need for fragmented data across multiple SaaS chains or middleware.

How it works

OB1 is built as an infrastructure layer consisting of a database with vector search capabilities and an open protocol. It uses a combination of Supabase Edge Functions, a PostgreSQL database with pgvector, and Model Context Protocol (MCP) servers to allow various AI clients to plug into the same data source. Users can capture information via integrations like Slack or Discord, import data from various platforms (ChatGPT, Obsidian, X/Twitter), and extend the system with custom schemas and agent skills.

Who it’s for

It is designed for power users of AI tools who want a centralized, and self-hosted memory system that their AI agents and assistants can recall and write back to, regardless of which specific LLM or AI tool they are currently using.

Highlights

• Unified Memory: A single database that serves as the persistent memory for multiple AI tools simultaneously.
• Extensible Architecture: Includes a library of "extensions" (e.g., Household Knowledge Base, Professional CRM) and "recipes" for importing data from diverse sources.
• MCP Integration: Uses the Model Context Protocol to connect to clients like Claude Desktop and ChatGPT.
• Open Protocol: Avoids vendor lock-in by using an open protocol and providing options for self-hosting (e.g., via Kubernetes).
• Agent Skills: Provides plain-text skill packs that can be dropped into AI clients to teach them specific workflows like research synthesis or meeting synthesis.