Canonical: https://molo17.com/ai-hub/enterprise-brain/ownership-and-privacy/ Markdown mirror: https://molo17.com/ai-hub/enterprise-brain/ownership-and-privacy/index.md Title: Ownership and privacy · Gluesync Enterprise Brain · MOLO17 Description: The Gluesync Enterprise Brain is self-hosted: a dedicated database on your infrastructure, your own models, memory shared by default and private on request, never result rows, never secrets, never SaaS. [Gluesync Enterprise Brain](/ai-hub/enterprise-brain/) · Own # You own it. You control it. It is not SaaS. Most AI assistants would like you to send your schema to their cloud and trust the invoice. The Enterprise brain is a database file beside your Core Hub data, on your infrastructure, fed by models from your own vault. It remembers where answers live, never the answers themselves, and it forgets when you tell it to. [Book a demo](/book-a-demo/) [Read the docs](https://docs.molo17.com/gluesync/latest/gs-modules/enterprise-brain.html) 1. 01 A dedicated brain database beside gluesync.db 2. 02 Your models, your keys, your permissions 3. 03 Memory of routes and procedures, never rows or secrets How it works ## Where the brain lives and what it keeps 1. 01 Location ### gluesync-ai-brain.db, on your host The schema graph, agent memory, conversations, skill attachments, and approved remote tools live in a dedicated SQLite database next to your Core Hub data. It is part of your backups and support snapshots, and it never leaves your infrastructure. 2. 02 Content ### Location and procedure, never data A memory says which tables answered which question and how. It never stores result rows, and password, token, secret, and API key assignments are redacted before anything is written. Telemetry carries identifiers and counts only. 3. 03 Access ### The caller's role, every time Conversations are private to their owner. Memory is partitioned by agent and visible under the same Core Hub RBAC that governs the UI. Private memories belong to one person and cannot be read or forgotten by anyone else. Capabilities ## Sharing, search, and models are choices, not defaults you inherit A company brain only works if people trust what it remembers and where it looks. Every knob here is visible in AI Studio and documented. Sharing ### Shared by default, private on request The point of a company brain is that the second person benefits from the first. Memory is shared across everyone who uses an agent unless the person asked otherwise. - Say "keep this private" or "private memory" and the memory stays yours - Forget a shared memory, or your own private one, from the API or the UI - Conversations are always owner-private; Super admins keep their existing view Search ### Keyword always on, embeddings suggested Search mode is a setting you see, not an add-on you buy. Keyword search over names, columns, and keys costs nothing and works with any chat model. - Embeddings are suggested when a model with the embeddings capability exists, unavailable otherwise - The embedding model is pinned with the brain; changing it rebuilds the vectors - Switching back to keyword stops embedding calls; the chat model is independent Models ### Your vault, your egress Chat and embedding models come from the providers you registered in Settings → LLM providers: OpenAI, Anthropic, Azure OpenAI, OpenAI-compatible endpoints, or a local Ollama box. - Credentials stay encrypted in Core Hub and never proxy through MOLO17 - Budgets, scoped keys, rate limits, and audit apply to brain-backed answers too - Route everything to a local model and the whole loop stays on-premises FAQ ## Ownership and privacy: questions we hear Is any part of the Enterprise brain hosted by MOLO17? No. The brain is a database inside your Core Hub deployment. MOLO17 never receives your schema, your memories, or your conversations. Gluesync Connect, our hosted operations console, is a separate product with its own read-only rules. What exactly is written to the brain after a question? The question text (redacted), the names of the tools and tables the agent used, and a short procedural note. Result rows are never stored, and secrets are stripped before writing. Memory text is capped at 2,000 characters. Can I turn embeddings off after trying them? Yes. Switch the search mode back to keyword and Core Hub stops calling the embedding model. Stored vectors are cleared, and keyword search over the same graph keeps working. We already run a 2.3 preview. What happens to existing conversations? Nothing is migrated and nothing is deleted. Earlier conversations stay in gluesync.db as they are; new conversations are written only to the brain database. Enterprise brain ## Keep exploring [Overview **Gluesync Enterprise Brain** The foundation for every AI tool your company uses, built on the databases Gluesync already integrates. Explore →](/ai-hub/enterprise-brain/) [Connect **Connect your AI tools** Point Claude, ChatGPT, Grok, Meta Muse, Hermes Agent, OpenClaw, or Cursor at Core Hub and let them ask the brain under the caller's permissions. Explore →](/ai-hub/enterprise-brain/connect-your-ai-tools/) [Extend **Skills and company tools** Agents publish the procedures they discover as skills, and approved MCP servers such as GitLab or Notion become tool bundles you govern. Explore →](/ai-hub/enterprise-brain/skills-and-company-tools/) Gluesync Enterprise Brain ## Yours, under your own rules. Bring the model you trust, keep the schema where it belongs, and give every AI tool in the company one brain that answers without ever leaving your building. [Book a demo](/book-a-demo/) [Get Gluesync](/get-gluesync/)