Most personal-knowledge tools return a list of pages; GBrain turns those pages into an actual answer. It runs a continuous ingestion + enrichment loop that extracts entities, auto-links pages into a typed knowledge graph, and synthesizes well‑cited prose with explicit gap analysis — so an agent can prepare for meetings, fetch a company's context, or give a coding agent memory-aware answers without re-reading hundreds of notes.
What Sets It Apart
- Synthesis-first workflow: instead of surfacing source snippets, GBrain composes a single, cited answer plus an explicit “what the brain doesn’t know” note. That gap analysis is the product hook that changes how teams rely on their memory store.
- Self-wiring knowledge graph: entity extraction on every write creates typed edges (attended, works_at, invested_in, etc.) without LLM calls, enabling multi-hop queries and graph-boosted retrieval that improves named-entity recall vs vector-only RAG.
- 24/7 dream cycle & enrichment: cron-driven jobs dedupe people, fix citations, and consolidate memory overnight so the brain continually improves and surfaces stale claims.
- Integrations & agent-first UX: out-of-the-box MCP connectors (Claude Code, Codex, HTTP/OAuth server), a CLI for local brains (PGLite) and scale paths using Postgres+pgvector for team deployments.
Who It's For & Tradeoffs
Great fit if your team needs an auditable, shared institutional memory for agents and humans — e.g., meeting prep, investor/founder CRM, or giving coding agents persistent context. It’s especially useful when you want precise citations and explicit notes about stale or missing information. Look elsewhere if you need a zero-ops drop-in hosted memory (GBrain expects data ingestion and some operational setup), if your data governance prohibits centralized indexing, or if you only need lightweight keyword search — GBrain’s value increases with volume and structured inputs (notes, emails, meeting transcripts).
Where It Fits
Think of GBrain as the layer between hybrid retrieval and the agent: it combines BM25/vector retrieval, graph signals, and a synthesis layer so agents stop being “amnesiac” about non-code context. For single-file/document lookups a simpler vector store may suffice; for company-scale memory with multi-source reasoning, GBrain adds measurable value.
