Built-in Rix AI is the zero-setup way to use intelligence in Rixoffice. You sign in, and every composer in every app just works — no API keys, no endpoints, no provider billing to manage. This page explains what is behind that default, what it costs, when it is the right choice and how to move beyond it if your usage changes.
What Rix AI actually is
Behind the scenes, built-in Rix AI is a proxy service operated by Rixcart. Your requests from the desktop app go to the proxy, which serves a curated set of frontier models:
- Claude Opus, Sonnet and Haiku for long-document reasoning, everyday writing and fast responses.
- GPT-5.2 for general reasoning and generation.
- Gemini 3 for fast multimodal work.
The point of the proxy is convenience. Rixcart maintains the provider relationships and model access, so the model list can improve over time without you touching a setting. You always talk to the same composer; the proxy decides the plumbing. When a better model appears behind it, you benefit without a migration, and when a provider has an outage, the service side handles it rather than your afternoon.
Why start with the built-in option
Three practical reasons:
- No configuration. Sign in and go. There are no keys to paste, no base URLs to test, nothing to break when a provider rotates credentials.
- One meter. Usage is billed in plan credits rather than provider invoices. One AI action is one credit, so you always know what a request costs. The details live in AI credits explained.
- Identical behavior. The same agents, skills and block-level editing run regardless of provider. Starting built-in and switching later changes nothing about how you work.
Getting started
- Sign in with your Rixcart account from the Rixoffice home screen.
- Confirm a plan is active — every subscription plan includes built-in Rix AI with a monthly credit allowance.
- Open any file and the composer, then type a plain-language request.
- Watch the edit land block by block in real time, and accept or refine it.
- Track consumption on the credit usage page or in Settings.
That is the whole onboarding. The first request you make is the first test of the setup, and there is nothing else to configure — which is exactly why this provider is the default.
What your credits pay for
On built-in Rix AI, credits meter the AI actions you take: edits, rewrites, summaries, chat answers and analysis are one credit each. A few flows are metered differently — PDF-to-Word cloud conversion costs 5 credits, and AI deck pages beyond your plan's included allowance cost 1 credit per 3 pages. RixSheets AI usage is unlimited pages on all plans.
Because credits are per action rather than per token, a long rewrite costs the same as a short one. That predictability is the main reason many subscribers stay on the built-in provider even after learning about the alternatives: a monthly budget of actions is easier to reason about than a provider invoice whose line items change with model pricing. It also removes the temptation to hoard usage — when every action costs one credit, you ask freely instead of rationing questions.
The proxy is also where quality control happens. Model availability is curated rather than dumped on you: you get a small set of strong options suited to office work instead of a catalog of hundreds of models and the burden of choosing between them for every request.
When to switch providers
Built-in Rix AI is designed to be the easy default, not a ceiling. Consider switching when:
- Volume outgrows credits. If you run hundreds of actions a month, your own API key on a provider you choose is usually cheaper. That is the BYO API keys route, available on the Pro plan and above.
- You need offline AI. A flight, a secure facility, a cabin — Ollama local models keep every AI feature working with no internet at all.
- You want a specific model. If your team standardizes on one provider's models, connecting a key puts that model behind the same composer.
Switching takes about a minute in Settings, and you can move back and forth whenever it suits the task. Nothing is lost in the transition: chat history, per-app skills and agent behavior stay the same. The AI provider setup guide covers all options side by side.
Privacy on the built-in path
Using the proxy means the text involved in a request passes through Rixcart's service to reach the model. Rixoffice keeps this as narrow as possible: your documents are processed locally, and cloud AI only receives the text you ask about, not your whole file. The full model is described in AI privacy. If even that is more than your work allows, the Ollama route keeps every byte on your machine — same composer, same agents, zero network.
A useful way to think about the trust model: with built-in Rix AI you are trusting Rixcart's pipeline in addition to the model providers; with BYO keys you are trusting only the provider you chose. Neither is wrong, but regulated industries and confidential work tend to prefer the shorter chain, which is exactly why the settings make it a toggle rather than an upgrade.
What to explore next
- AI credits explained — exactly what each action costs.
- AI provider setup — the full provider list and how to switch.
- AI privacy — what leaves your machine and what never does.