Grounded answers about the platform — no hype.
Hi, I'm Dewey — laibrary's website assistant. Ask me anything about laibrary. I answer only from what our team has published, and I point you to a human for anything I can't cover.
Try asking
AI-generated — don't enter confidential information. Privacy
Most organizations discover their AI footprint after something goes wrong. Governing it means holding a current, owned inventory — and checking every action against your rules before it happens, not reporting on it afterwards. Every regime that binds you, from the EU AI Act to a US state statute to your sector regulator, starts by asking for that inventory. It is the part that does not get easier with delay.
Every agent — whatever framework built it, whichever vendor runs it — catalogued with an owner, a purpose, a risk tier and the systems it is allowed to reach. Nothing runs off the books, and the register is the one artefact every framework asks for first.
Twelve built-in policies check each call an agent makes — budget, compliance state, data freshness, revoked access, prompt injection in either direction — and warn or block per policy.
Write your own rules in a versioned DSL, roll them back like any other artefact, and route matches into an approval queue instead of a refusal.
Find the AI nobody registered — from identity sign-in logs, CASB feeds, your Microsoft 365 and Copilot audit log, or an opt-in browser extension.
Operators, admins, auditors and the AI governance lead each see the same record rendered for the job they are doing, rather than one screen with everything on it.
Every state-changing action is chained to the one before it — alter a single entry and the chain no longer verifies. Your tenant's data is isolated at the database row and held in the region you choose.
Agents arrive by onboarding wizard, SDK, vendor connect or shadow-AI triage. Each gets an owner and a risk tier before it is allowed to act — and the industry pack you chose at onboarding sets the approval chain, with the roles your sector actually uses.
Calls route through the policy gateway. Failing checks warn, block, or pause into an approval queue, depending on how you set the rule.
The resulting record is the evidence — hash-chained, verifiable in-app, and assembled into audit packs without a separate collection exercise.
Tell us your industry, where you and your customers are, and how far along you are. We will show you the record your tenant would produce and the obligations it maps to.