Home / Projects / Open Brain

Open Brain: Unified Context Layer

Cross-Tool AI Supabase MCP pgvector Semantic Search

Overview

Open Brain is a unified context layer that works across every AI tool. It's a tool-independent knowledge base built on Supabase and pgvector, connected via MCP (Model Context Protocol). Switch the tool, keep the memory.

Every learning captured through the Second Brain note-taking workflow feeds into this layer. Whether you're using Claude, ChatGPT, Gemini, Cursor, or any future AI agent, they all read from and write to the same knowledge base.

The Problem

Every AI tool today operates in its own silo. The context you build with Claude stays in Claude. Your ChatGPT history lives in ChatGPT. Switch tools, and you start from scratch.

There's no portability. Your knowledge, preferences, and identity are locked inside each vendor's platform. If a better tool comes along tomorrow, you can't take your context with you.

How It All Connects

AI tools connect to one shared knowledge base through MCP. Switch the tool, keep the memory.

Open Brain architecture diagram showing AI tools (Claude, ChatGPT, Gemini, Cursor, Future Agents) connecting via MCP Protocol to a Unified Context Layer on Supabase with tables for Identity, Notes, YouTube transcripts, Articles, Memory, and Configuration

Any AI tool that supports MCP can plug into Open Brain instantly. No migration, no re-training: just connect and your entire knowledge base is available. Every learning captured once is reusable across every tool, forever.

The three layers work as progressive disclosure:

  • Layer 1: BIOS A small prompt in each tool's settings. When you open a new chat with Claude, ChatGPT, or any AI tool, it already knows Open Brain exists and has high-level context about your knowledge base.
  • Layer 2: OS (brain_config) Tells the AI how to operate with Open Brain: table structures, save rules for new data, and search instructions so it knows how to query your knowledge when you ask.
  • Layer 3: Tables The actual Supabase tables where everything lives:
    • Identity: Background, preferences, and personal context
    • Knowledge Notes: Learnings captured through Second Brain
    • YouTube Transcripts: Insights from videos and podcasts
    • Articles: Key takeaways from long-form reads
    • Memory: Persistent facts that carry across sessions
    • Configuration: Shared behavior rules and response style

How MCP Connects Everything

MCP (Model Context Protocol) is the bridge. A Supabase MCP Server provides account-level access to the knowledge base. Any AI tool that supports MCP can read from and write to the same set of tables.

This means adding a new AI tool to the system is simple: paste the BIOS prompt into its settings, connect the MCP server, and it immediately has access to your full context. No migration, no re-training, no data export.

The Core Idea

  • Your context should be yours, not the tool's AI tools come and go. Your knowledge and identity should outlast any single product.
  • One config to rule them all brain_config acts as a shared OS. Change behavior in one place, and every connected tool picks it up.
  • Future-proof by design Any tool that supports MCP can plug in. No custom integrations needed per tool.