Pricing
Pricing that scales with your usage
Every plan includes the whole product — assistants, knowledge, flows, inbox and insights — for as many members as you like. What changes is how much AI work the plan funds each month, and how deeply Ciele plugs into your existing systems.
Answers a month
Go funds about that much answering — it is marked below. Crawling and indexing draw on the same allowance, so a heavy indexing month leaves less for answers.
What a plan pays for
Members are unlimited on every plan, and visitors chatting with a published widget are never counted. What a plan meters is the AI work the platform funds on your behalf: answering questions, crawling your sites and indexing your documents. Each plan below states that allowance as the volumes it covers, and your Usage page shows exactly where you are against it.
Answer volumes are quoted on Gemini 3.1 Flash Lite, the lightest model on the platform. Which model your assistants run is your choice and it moves this a lot: on Claude Sonnet 5 — the usual step up when you want a stronger answer — one answer costs roughly 27× more, so the same allowance covers proportionally fewer.
Need more than the top plan funds, a multi-campus rollout, or procurement requirements? Talk to sales and we will quote it properly.
Compare every plan
Ciele is open source under the AGPL, so running it yourself is always an option — you bring the infrastructure and the model account, and you keep the whole product.
Install the self-hosted edition
Clone the repository and it runs. With no database configured the console starts on an in-memory demo dataset, so you can click through the whole product before wiring anything up.
git clone https://github.com/MattiaIppoliti/ciele.git && cd ciele
pnpm install
pnpm devWhat you provide
Node.js and pnpm
The apps are standard Next.js applications; anywhere that runs Node can host them.
A Postgres database
Supabase provides the database, authentication and row-level security. Docker runs the whole stack locally.
Your own model keys
Any supported provider, or point at an OpenAI-compatible endpoint to run the models locally too.
AGPL-3.0. Hosting, upgrades, backups, plan billing and support are the managed edition’s job — everything else is in the repository.