Mother Brain: Multi-dimensional memory for every app, AI, and human

We built Mother Brain because we were tired of re-explaining our codebase to AI assistants. Every. Single. Session. We'd spend 20 minutes giving Cursor context about our architecture — file relationships, design decisions, why we chose one pattern over another. Then we'd close the chat, come back the next day, and start from absolute zero. Multiply that across weeks, projects, and team members, and it's hours of pure waste. The breaking point came when we were working on a complex feature and realized our AI knew less about our codebase than it did the day before — because we'd started a fresh chat. That's when we thought: *what if the AI never forgot?* We wanted three things that no existing tool delivered: 1. Real persistence. Not just "remember the last file" — we wanted semantic search across our entire codebase, every past conversation, every architectural decision, even our git history. Something we could ask "where did we decide to use this pattern?" and get an instant answer. 2. Local-first. Our code stays on our machine. We didn't want to upload our codebase to a cloud service. Mother Brain ships with an embedded PostgreSQL 17 database with pgvector — zero server setup, zero cloud dependency. 3. Works with everything. We use Zed, sometimes Cursor, sometimes VS Code. We didn't want another walled garden. So we built Mother Brain as an MCP (Model Context Protocol) server — it connects to any MCP-compatible tool. We're also using it with OpenClaw and Hermes agents for multi-agent workflows. The thing we're most proud of technically is VMVA (Vertical Multi-Vector Architecture) — our 3-layer code search pipeline that uses three different Voyage AI embedding models in cascade: fast file filtering → LLM-generated file summaries → deep code chunk retrieval. It finds the exact code you need in milliseconds. We're pricing it at $25/year — that's less than $3/month. No tiers, no upsells, no "enterprise" paywall. We wanted it to be a no-brainer. Beta testers get a lifetime license for $25 one-time. We'd love your feedback — especially on the MCP integration and the code indexing. Ask us anything about the architecture, the embedding pipeline, or how it fits into your workflow. 🧠