ob1

Open Brain (OB1) is an infrastructure project for personal AI memory. Rather than a notes app, it is described as a database with vector search plus an open protocol, so that multiple AI tools — Claude, ChatGPT, Cursor, Claude Code and others — can read and write to the same persistent store of a user's thoughts. Core components described in the README: - A database (PostgreSQL with pgvector) holding "thoughts" - An AI gateway - A chat capture channel (Slack, with a Discord alternative) - An MCP server exposing the brain to AI clients The project positions itself as an alternative to middleware and SaaS chains, with a self-hosted Kubernetes deployment option that uses PostgreSQL + pgvector and does not require Supabase. Getting started material includes a setup guide (roughly 45 minutes, no coding experience claimed), a video walkthrough, an AI-assisted setup path for tools like Cursor or Claude Code, a FAQ, and a set of companion prompts for migrating memories and building a capture habit. Dedicated assistant configurations are offered for Claude, ChatGPT and Gemini. Repository structure: - Extensions: a curated learning path of six projects (household knowledge base, home maintenance tracker, family calendar, meal planning, professional CRM, job hunt pipeline), ordered by difficulty. - Primitives: reusable concepts such as deploying an edge function, remote MCP connections, troubleshooting, row level security, and shared MCP servers. - Recipes: community-contributed importers and workflows, including imports from ChatGPT, Perplexity, Obsidian, X/Twitter, Instagram, Google Takeout, Grok, Blogger/Journals and Gmail, plus tools like auto-capture, idea mining, schema-aware routing, fingerprint deduplication, source filtering, daily digest, and several agent-memory and research workflows. - Skills: plain-text skill packs for AI clients, covering auto-capture, competitive analysis, financial model review, deal memo drafting, research synthesis, meeting synthesis, idea mining, skill generation, work operating model elicitation, and world-model diagnostics. - Dashboards: two frontend templates (SvelteKit and Next.js) intended to be hosted on Vercel or Netlify against a Supabase backend. - Integrations: Kubernetes deployment, an agent memory API, an OpenClaw plugin, and Slack and Discord capture bots. - Schemas: database extensions such as agent memory sidecars for provenance, review, use policy, relations, recall traces and audit. Contribution process: extensions are curated and require maintainer discussion; primitives should be referenced by at least two extensions; recipes, schemas, dashboards, integrations and skills are open to community contributions. Pull requests pass an automated review agent checking structure, secrets, SQL safety, dependencies and documentation, followed by human review. Contributions require a README and a metadata.json. Community channels are a Discord server and a Substack newsletter. The project is created by Nate B. Jones and maintained by a small team. License is FSL-1.1-MIT.