Private AI, without the trade-off
Most enterprise AI asks you to choose between capability and control. Send your documents to a cloud provider and hope the contract holds, or keep them in-house and go without. Noema Teams refuses that trade. Large Language Models run on each employee's iPhone, iPad, or Mac — offline-capable and private by default — while a company workspace adds governed datasets, role-based permissions, and admin control over exactly what each person can use. Datasets live on company devices, retrieval happens on-device, and policy decides what every role can touch.
How it works
You create a workspace and receive a company code. Employees open Noema, go to Settings → Enterprise, enter the code and their work email, and verify with a one-time code. Their device registers and fetches a policy snapshot — which datasets, tools, model formats, specific models, and remote backends their roles allow. The app caches that policy and enforces it locally, even when fully offline.
Governed datasets, indexed on device
Upload handbooks, research, contracts, or course material to your workspace and assign them to roles. Permitted employees see them appear in Noema as company datasets — indexed and searched entirely on device through the same Retrieval-Augmented Generation pipeline Noema users already know. Files are hash-verified on download, stored in app-private storage, and removed automatically when access is revoked.
Roles, not rules-of-thumb
Default roles — Owner, Admin, Member, Research, Engineering, Finance, Legal, Teacher, Student, Volunteer — are ready out of the box and fully customizable. Per role you decide which datasets are visible, which tools run, which model formats and specific models can load, which remote backends are reachable, whether remote inference is allowed at all, and whether the device must operate off-grid.
Enforced in the engine, not the interface
Policy checks sit inside Noema's execution paths. A disallowed tool doesn't run. A disallowed model doesn't load. A disallowed backend can't be called. If a role requires off-grid operation, the app's network kill switch engages — the same one consumer Noema uses — allowing only policy sync with your workspace. Expired policies degrade safely: cached restrictions stay in force and company datasets go dark until the device syncs again.
Identity, access, and a console that sees nothing it shouldn't
Membership is tied to verified work email. Invite employees directly, restrict joining to your email domain, or require admin approval for every request. Each device is individually registered and individually revocable from the console. Everything is managed from the website — Overview, Employees, Roles, Datasets, Models & Tools, Invites, Devices, Settings, and an activity stream of administrative events. Telemetry stops there: Noema never sees your documents, prompts, or answers.
Who it's for
Companies with internal knowledge that can't leave the building. Schools giving students a safe, governed assistant. Research groups working with embargoed data. Clinics, law practices, campaigns, and nonprofits — teams that need private AI without losing control of internal knowledge or paying per token. Noema Teams targets iOS, iPadOS, and macOS first, with Apple Silicon recommended.
Free during Early Access
Noema Teams is free during Early Access. Pricing may be introduced after General Availability with at least 60 days' notice and a reasonable transition period for existing workspaces. On the roadmap: signed policy verification, OIDC single sign-on, encrypted dataset bundles, ingestion connectors, and MDM distribution.