Concerto Local · 999 €/month

Total data sovereignty

Shara on your own servers, with every Concerto model running on your hardware. It also brings 60 M STU a month of cloud quota for the tasks you want to hand to Amadeus: it is a switch, and with it off nothing leaves your network. For regulated sectors, or contracts with strict sovereignty clauses. It starts at six months: we install it on your own infrastructure, on site or remotely, and that setup does not pay for itself in a single month.

Sovereignty is your call

With the cloud quota switched off, not one task leaves your network.

All on your hardware

The director and the Concerto agents run on your own server.

All-inclusive

No overages, and 60 M cloud STU a month inside the price.

999 €
Per month, all-inclusive
Six-month minimum term
3
Concerto tiers
Matched to your server
30 days
Grace period offline
Everything keeps running locally
60 M
Cloud STU a month
Optional: you switch them on

The difference

Cloud vs on-premise

The cloud plans live on our European infrastructure; Concerto Local lives on yours, and the only thing that leaves your network is what you choose to send.

Cloud (First/Pro/Max)

Data inside the European Union
Set up in hours, no hardware needed
Premium models (Symphony class)
Data passes through LLM providers (under SCCs)

Concerto Local (on-premise)

Your data and the Concerto models live on your own infrastructure
Every model, the director and the agents, runs on your hardware
60 M STU a month of cloud quota, optional: you switch it on or off
All-inclusive: no overages and no usage bills
A six-month minimum term: it is installed on your infrastructure, and that does not pay for itself in less
Email support, with a person behind it, in European business hours

The cloud quota

You decide what leaves your network

Concerto Local includes 60 M STU a month of cloud quota, to hand Amadeus the tasks you want to hand it. It is a switch in the admin console, not a pipe that is always open: while it is off, no task leaves your server.

With the switch off

Everything happens inside your network: the Concerto models, Amadeus, the 20 agents and your data. No task travels anywhere, and not a single STU of the cloud quota is spent.

  • Amadeus and the agents run on the Concerto model on your hardware
  • No task and no data of yours leaves your network
  • Licence validation is the only thing that talks to the outside world

With the switch on

You can hand Amadeus in the cloud the tasks that need a more capable model than your server can hold (strategy, long-form analysis) without buying more hardware. Those tasks go out and only those, and only when somebody sends them.

  • 60 M STU a month, included in the price of the plan
  • No overage: the quota is fixed and does not carry over to the next month
  • Inference is processed inside the European Union, with no international transfer

Tiers

Three Concerto tiers, matched to your server

Concerto Local is deployed on your infrastructure. How powerful your server is decides how capable the local model can be. We help you pick the right tier for your volume and your budget.

Tier 1

Concerto Compact

Small server

An office of about five people, or a single department.

Tier 2

Concerto Standard

Mid-range server

A general-purpose workhorse for the whole team.

Tier 3

Concerto Extended

Large / high-capacity server

Maximum throughput: the best local model available.

Technical requirements

What hardware each tier needs

These figures are indicative: we size the deployment with you, against your real volume.

Tier Compact Standard Extended
Server Small serverMid-range serverLarge / high-capacity server
RAM 64 GB of RAM (ECC)128 GB of RAM (ECC)256 GB of RAM or more
GPU / VRAM One 24 GB GPU (RTX 4090 or A5000)Two 40 GB GPUs (A100 or L40S)Four 80 GB GPUs (A100 or H100)
Who it is for An office of about five people, or a single department.A general-purpose workhorse for the whole team.Maximum throughput: the best local model available.

We also support private-cloud deployments (AWS, Azure, GCP) with dedicated GPU instances inside your own tenant.

How the licence works

It runs on your network, even offline

With the cloud quota switched off, licence validation is the only thing that talks to the outside world. And even if that goes down, Shara keeps working for 30 days.

Licensing service

Your local instance validates its licence against our cloud service using Ed25519 signatures. The signature carries the expiry date, the plan and the agents you have switched on.

  • Ed25519 signature on the edwards25519 curve
  • Configurable validation interval
  • Automatic renewal, with confirmation

30-day grace period

If you lose contact with the licensing service, Shara carries on as normal for 30 days: because every Concerto model runs on your hardware, all that pauses is licence validation, telemetry, billing and, if you had it switched on, the cloud quota.

  • Every Concerto model keeps running locally
  • No telemetry and no billing sent
  • Automatic reactivation once the connection is back

Regulated sector? Let’s start with a pilot

We design a 30 to 90-day pilot to prove Concerto Local on your infrastructure before anything is signed.