GBrain iconGBrain

open-source Free Star23k

The brain layer your AI agent has been missing - synthesis, graph traversal, and gap analysis for intelligent agents

20.8K GitHub Stars
43 Curated Skills
MIT License

Overview

GBrain is an opinionated agent brain layer that goes beyond simple RAG to provide synthesis, self-wiring knowledge graphs, and gap analysis. Unlike traditional search that returns raw pages, GBrain generates well-cited prose answers while explicitly identifying what the brain doesn't yet know. It operates as a daemon that ingests meetings, emails, tweets, voice calls, and ideas, automatically linking entities and enriching data. Built with TypeScript/Bun and supporting both local (PGLite) and Postgres deployments, it's designed for developers who want their agents to truly understand and reason over accumulated knowledge.

The Verdict

Who Should Use GBrain?

Best For

  • Developers building personal knowledge agents
  • Teams wanting synthesis, not just retrieval
  • Local-first, privacy-conscious deployments
  • MCP-compatible agent architectures
  • Projects needing automatic entity linking

Not Ideal For

  • Simple chatbot memory needs
  • Teams wanting managed cloud service
  • Non-TypeScript/Bun environments
  • Projects needing minimal setup

What's Great

  • Synthesis layer generates cited answers, not just chunks
  • Self-wiring knowledge graph with typed edges
  • Gap analysis identifies what the brain doesn't know
  • Hybrid search: vector + BM25 + graph signals
  • 43 curated skills for signal capture and enrichment
  • Local-first with PGLite (WASM-based Postgres)
  • MCP support with stdio and HTTP modes
  • 16 embedding provider options including local

Watch Out For

  • Opinionated architecture may not fit all use cases
  • Requires TypeScript/Bun ecosystem familiarity
  • No managed cloud option (self-hosted only)
  • Steeper learning curve than simpler memory solutions
  • Third-party API costs for embedding/LLM providers

Pricing

View all features & details

Core Features

  • Hybrid search (vector + BM25 + graph)
  • Self-wiring knowledge graph
  • Typed edges (attended, works_at, invested_in, founded, advises)
  • Synthesis with citations
  • Gap analysis
  • 43 curated skills
  • Schema packs (15-type taxonomy)
  • Job queue with Postgres-native architecture

Tech Stack

  • TypeScript/Bun runtime
  • PGLite (local WASM) or Postgres
  • pgvector for embeddings
  • MCP (stdio + HTTP + OAuth 2.1)

Embedding Providers

  • OpenAI
  • Voyage
  • ZeroEntropy
  • Ollama (local)
  • llama-server (local)
  • 16 options total

Use Cases

  • Personal knowledge assistant
  • Meeting/email synthesis
  • Research agent brain
  • Entity relationship tracking
  • Multi-brain federation

How It Compares

Feature GBrain Mem0 MemClaw
GitHub Stars 20.8K 57K+ N/A (Commercial)
Architecture Self-wiring graph + synthesis Multi-level memory Enterprise shared memory
Gap Analysis Yes, built-in No No
Synthesis Cited prose answers Memory retrieval Memory retrieval
Managed Cloud No Yes Yes
Open Source MIT Apache 2.0 No
Best For Personal/team knowledge synthesis Personalized AI assistants Enterprise multi-agent fleets

User Reviews

Loading reviews...