The memory layer AI vendors forgot to build

What it does

Open Brain is a Supabase-based stack—PostgreSQL with vector search, an AI gateway, and an MCP server—that acts as a central memory store for your personal data. Instead of siloed context windows, every AI client you plug in (Claude, ChatGPT, Cursor, and whatever ships next month) queries the same persistent database of your thoughts, schedule, contacts, and imported digital history. The project ships with a curriculum of extensions that gradually turn the database into a household knowledge base, CRM, meal planner, and job-hunt tracker.

The interesting bit

The real architecture choice is betting on the Model Context Protocol (MCP) as the universal plug. Rather than building bespoke integrations, Open Brain exposes your memory as an MCP server, so any client that speaks the protocol can read and write your brain state without middleware or SaaS chains. It is essentially personal infrastructure dressed up as a very opinionated database schema.

Key highlights

• Vector-searchable PostgreSQL backend with community recipes for importing ChatGPT logs, Obsidian vaults, Gmail, X/Twitter, Instagram, Google Takeout, and more.
• Extension track that compounds: household knowledge → home maintenance → family calendar → meal planning → professional CRM → job hunt pipeline.
• Active community contributions including cost-optimized edge functions (73% invocation reduction), provenance tracking, and a Next.js dashboard.
• Row-level security primitives for multi-user isolation when you inevitably share part of your brain with family or coworkers.
• Dedicated AI assistants (Claude Skill, ChatGPT Custom GPT, Gemini GEM) that know the system well enough to help you build it.

Verdict

Worth exploring if you model-hop between Claude, ChatGPT, and Cursor and want them to share one long-term memory. Skip it if you are looking for a simple note-taking app; this is a database that wants to run your household, your CRM, and your job hunt.