AI, on your terms

No AI inside. All of it available.

Orma ships no chatbot and no model. Nothing of ours reads your trips, nothing sends them away to be summarised, and nothing trains on them. But if you already use an assistant, you can open a door — one you scope, watch, and close.

What is inside Orma

Nothing. There is no chatbot, no model of Orma's own, and nothing that reads your trips. Orma never sends your trips anywhere to be summarised, and never trains any model on them — not its own, not anyone else's.

That is a design decision, not a roadmap gap. A planning tool should be a quiet surface for your own thinking, and the trips on it are yours.

What you may bring

Orma exposes a Model Context Protocol (MCP) server, so an assistant you already use — Claude, ChatGPT, or any MCP client — can work with your trips if you choose to connect it. This is off by default, and connecting is a deliberate act in your account settings.

The assistant is yours: it runs under your account and your agreement with its provider. Orma is not affiliated with, endorsed by, or partnered with any assistant provider. The connection feature is in alpha.

Access is scoped and revocable

Four modes: read-only or read-and-write, across your whole account or only the trips you pick.

  • Read-only means read-only. Such a connection writes nothing at all, not even an advisory note.
  • Scope is resolved live, so a change takes effect on the assistant's next request.
  • Every connection carries an activity log.
  • Revoking is one switch.

Changes wait for you

An assistant cannot edit a trip's contents. Edits are held as proposals that do nothing until you open Orma, read them and approve them. Nothing is deleted without you either — not an event, not a scenario, not the trip. Because deleting a scenario or a trip cannot be undone, approving one asks you to confirm a second time, and only a trip's owner can approve deleting it.

What does apply directly

Under read-and-write access an assistant can, without a proposal: create a trip; create a scenario (which can copy an existing scenario's events into it); switch which scenario is active; leave advisory notes on a trip beside Orma's own checks; and make any wishlist change — including deleting a wishlist entry, which is not yet held for review, though it goes to "Recently deleted" and can be restored for 30 days.

Nothing on that list loses work you can't get back. If you want a harder guarantee, grant read-only or per-trip access.

Your data and the provider

If you do connect an assistant, your trip data goes to that provider when you use it, under their privacy policy rather than Orma's — and some providers train on what you send unless you turn that off in their settings. That is the one place Orma's no-training promise cannot reach, which is why the door is closed until you open it.

Where this is written down

  • Terms of Use — section 7 sets out the AI position in full and enumerates exactly what an assistant may do.
  • Privacy Policy — section 7 covers the data that leaves Orma when you connect one.
  • FAQ — the short answers, including how to connect and disconnect.