GBrain
🔥 This release made it to our blog
GBrain: Garry Tan's Open-Source AI Agent Brain With Self-Wiring Knowledge Graph
GBrain is an open-source AI agent memory system built by Garry Tan, President and CEO of Y Combinator, to power his personal agents in production. It features a self-wiring typed knowledge graph, hybrid vector + graph search with P@5 of 49.1%, and a fully autonomous doctor command that optimizes brain health without human intervention.
### TL;DR
GBrain is an AI agent long-term memory and self-wiring knowledge graph system built by Garry Tan (YC President & CEO) to power his personal AI agents. It gives agents a persistent, searchable brain with hybrid vector+graph search, automatic entity extraction, self-healing citation fixing, and overnight memory consolidation — now open-sourced for any developer to use.
Key Insights & Metrics
Key Features
- Self-wiring knowledge graph — every page write auto-extracts typed entity links (attended, works_at, invested_in, founded, advises) with zero LLM calls; hybrid search combining vector embeddings, BM25, and graph traversal achieves P@5 49.1%, R@5 97.9% on 240-page benchmarks
- 43 scaffolded agent skills — signal detection, brain-ops, ingest, enrich, citation-fixer, daily-task-manager, cron-scheduler, eval framework; integrates as a skillpack into OpenClaw and Hermes agent platforms with 30-min setup
- Autonomous self-healing via gbrain doctor — one command computes a dependency-ordered remediation plan, submits jobs, re-checks score between steps, and refuses to exceed a user-defined cost cap; cron-driven autopilot runs unattended
Discussion
Sign in to leave a review
Reviews
No reviews yet. Be the first to review!