Frequently asked questions

STU (Shara Token Units) is the unit we measure your AI team’s consumption in. Every plan includes a monthly STU quota: First 50 M, Pro 140 M and Max 280 M. Concerto Local includes 60 M, which are its cloud quota: whatever runs on your own server spends no STU. Each conversation consumes STU according to its length and the model used, and you see it live in your dashboard.

You keep working, uninterrupted. The overage is billed at a flat 5 €/M STU (the same on every paid plan) and added to your next invoice. On top of that, if you use 80 % or more of your quota, the remainder rolls over to the following month, capped at 20 %. This applies to the cloud plans, not to Concerto Local.

First (129 €/month) includes Amadeus plus 5 essential agents, already with a per-agent IDENTITY.md (they learn your tone and your processes). Pro (349 €/month) adds 5 more specialists, a more capable model behind Amadeus and semantic memory across conversations. Max (690 €/month) adds the executive suite (strategy, engineering, product, communication, cybersecurity), the top-tier model and add-on agents you can switch on. Concerto Local (999 €/month) is the on-premise option.

Paying month to month, yes: the cloud plans (First, Pro, Max) have no minimum term. You can pick three periods (one month, six months or a year) and the two long ones are the ones that give you free months: one if you pay six, three if you pay the year, so the commitment is that period: it is charged up front and renews when it ends. You keep full ownership of your data and can export it in standard formats (JSON/CSV). Concerto Local starts at six months, because it is installed on your own infrastructure, and is governed by its own licence agreement.

Inference and data are processed entirely inside the European Union, on European sovereign infrastructure, with no international data transfer. We apply per-company isolation across several layers, AES-256 encryption at rest and TLS 1.3 in transit, MFA, continuous red-teaming (54 adversarial cases) and automatic spend kill-switches.

No. We do not train models on your data, and we work with opt-out enabled at the inference providers. Your data is yours: it is used to provide the service and for nothing else.

Yes. We have a signable DPA as data processor, automated data-subject rights, a record of processing activities and a product DPIA. We classify Shara as a limited-risk system under the EU AI Act, with full transparency and traceability. Data is hosted in the EU.

Concerto Local is Shara deployed on-premise: the app, the local Concerto engine and every agent run on your own server. It also includes 60 M STU a month of cloud quota, for the tasks you want to hand to Amadeus: it is optional, you switch it on and off from the admin console, and with it off nothing leaves your network. It is built for regulated sectors and contracts with sovereignty clauses. It includes Amadeus and 20 departmental agents (the 15 from Max plus the 5 add-ons at no extra cost) and a 30-day grace period.

Three, depending on your hardware: Concerto Compact for small servers (from 16 GB of RAM), Concerto Standard for machines with a GPU (around 24 GB of VRAM) and Concerto Extended for servers with a dedicated GPU, for maximum throughput. We help you pick the right tier during the pilot.

Support is us, by email, in European working hours, and a person reads it. Paid plans get priority in the queue. We do not publish response times or availability commitments: with no customers and no measured history we could not stand behind them, and we would rather not promise what we cannot yet guarantee. When we can, they will be published with their real figure. Setting it up is on you with First and Pro, and we walk you through it on Max and Concerto Local.

First: self-serve sign-up, running in 24 hours. Pro: self-serve, normally operational within a week. Max and Concerto Local: sign-up guided by a consultant, 2 to 4 weeks depending on the connectors and the IDENTITY.md files to train.