What is charged

Each kind of usage has its own price in the price list. This page explains what each one counts and when it is charged. The prices themselves are on the Prices page (/prices) and come from the billing.prices API procedure. Neither needs a sign-in. billing.prices returns the prices in effect now, price changes already scheduled with the time they take effect, and the VAT rate the prices include. A billing account’s Usage tab shows what each workspace used of each kind, and its Ledger tab every charge. See billing accounts for balances, members and price changes.

Who pays

Usage belongs to a workspace and is charged to the billing account that pays for that workspace at the time. After a workspace moves to another billing account, the current hour is still charged to the previous one. Usage outside any workspace is not charged.

When charges happen

• When it happens: assistant replies and web searches, document indexing, document text recognition, assistant sandbox time, email and SMS. Each is charged at the price in effect at that moment and appears in the ledger as its own entry.
• By the hour: endpoints, static VPN addresses, parameter requests, API requests, automation steps, stored readings, activity history, alarm log, files, map loads and address lookups. Each hour is charged shortly after it ends, at the prices in effect during that hour, as one ledger entry per workspace and kind of usage. A workspace that is deleted is charged for the rest of the hour it was deleted in.

Usage is counted for the day and hour it happened, not when it was charged.

A price per month is a price for 30 days: each hour is charged 1/720 of it. Every charge is rounded to the nearest millionth of a euro. Prices include VAT.

Devices and data

Usage	What counts	Price per
Endpoints	Every enabled endpoint, connected or not. Endpoints are checked every 15 minutes; one enabled at any check in an hour counts for that hour.	endpoint per month
Static VPN addresses	Every WireGuard endpoint, also while it is turned off or its workspace is suspended, since its address stays reserved. PIN devices connect through the VPN free of charge.	address per month
Parameter requests	Every parameter value Overvis reads from or writes to a device: background polling, live views, dashboards, automations and the API alike. Each parameter in a request counts once.	million requests
Stored readings	Raw readings kept in history, for every hour they are kept. Aggregated values are not counted. Deleting readings or shortening retention lowers it.	million readings per month
Activity history	Records in the workspace’s activity history, for every hour they are kept.	million records per month
Alarm log	Events in the workspace’s alarm log, for every hour they are kept.	million events per month
Files	Workspace files, each distinct content once, and the earlier versions the activity history keeps for checkpoints and reverts.	GB per month

A suspended workspace accrues no endpoint, stored reading, activity history, alarm log or file usage.

Every successful read is stored as a reading, whether background polling, a live view, a dashboard, an automation, or the API made it, and so is every reading recorded on an MC252. Each value read from a device is stored once, even when requests that arrive together share it. Readings stay until retention deletes them; with Keep everything, the default, they are never deleted. See Estimate a workspace’s monthly cost for a worked example.

Automation and API

Usage	What counts	Price per
API requests	Each API call that works in a workspace, counted with its procedure’s weight. A subscription or streamed call counts once. A device API call counts with its weight for every command sent to the device, also when the device then fails.	million weighted requests
Automation steps	Each step a stored automation run executed, failed runs included, and each check of an alarm rule.	million steps

API calls that touch no workspace are not counted. The parameter reads behind an API call count as parameter requests.

API request weights

Every API procedure has a fixed weight, listed with it in the API procedure reference. The cloud app, the local app, the CLI and your own integrations pay the same weight for a call.

• Scale: 1 is reading one workspace (workspace.get). Other weights are relative to it: about 1 for reading one item, 2 for a list, 3 for a change, and up to 10 or more for history queries, exports, bulk changes and whole-workspace operations. Weights are one of 0, 0.5, 1, 1.5, 2, 3, 5, 10, 20, 50.
• Worst case: a weight is set for the largest input the procedure accepts, such as the most items or the widest time range, and does not change with the size of a call. A call with many items costs the same as a call with one.
• Free (weight 0): hello, signing in and out and the session, help and discovery procedures, and every subscription, such as workspace changes, live values and notification streams. The parameter reads behind live values still count as parameter requests.
• Not counted: a call refused before it runs (invalid input, no access, rate limited) and a call that fails, unless it already did costly work, such as a device API command sent to the device. Work that has its own price, such as parameter requests, assistant replies or map loads, is not counted again in the weight.

In ledger entries and the price list, API requests are counted in half weighted requests, so a call of weight 1.5 counts 3 and the price applies per two million.

Assistant

Usage	What counts	Price
Assistant replies	Every model request of a reply, including the summaries that keep long chats within the model’s limits.	million input, output and cached tokens, per level
Assistant web search	Every web search the assistant makes, and the tokens of the model that runs it.	search, and million tokens
Assistant sandbox	The time the assistant’s sandbox for a workspace runs, by the second.	per hour
Document indexing	Preparing uploaded files so the assistant can search them.	million tokens indexed
Document text recognition	Every PDF page whose text is recognized.	per page

These prices follow the current prices of the model behind each assistant level, and of the models that run web search and indexing, and change with them; the prices page shows them. A model may charge more for a very long request. Searching the product documentation is free.

The Free assistant level is never charged, including a sandbox a Free chat started. Standard and Pro replies are charged to the workspace open while you chat, if you can view it; a reply with no such workspace is not charged. The chat shows what each reply was charged and nothing when it was not charged.

Maps and notifications

Usage	What counts	Price per
Map loads	Each map loaded.	thousand loads
Address lookups	Each address looked up for a map position.	thousand lookups
Email	Each notification email the mail provider accepted.	thousand messages
SMS	Each SMS part the SMS provider accepted. Up to 160 Latin characters (70 with other characters) is one part; a longer SMS is split into parts of 153 (67) and cut at two parts.	SMS part

Notifications are charged to the workspace they are about. In-app and Telegram notifications are free. Email and SMS that fail, are cancelled, or are stopped by the workspace’s monthly SMS limit are not charged, nor are messages Overvis sends on its own behalf, such as sign-in codes, suspension notices and announcements.

Estimate through the API

billing.estimate prices a usage profile at the current prices and needs no sign-in. Describe the devices in groups (how many, parameters each, reading period, retention, endpoints, connection), plus monthly amounts of anything else. For the ten meters of the worked example, kept for one year, the input is:

{
  "profile": {
    "deviceGroups": [
      {
        "name": "energy meters",
        "devices": 10,
        "parametersPerDevice": 5,
        "pollSeconds": 60,
        "retentionDays": 365,
        "endpoints": 1,
        "connection": "pin"
      }
    ]
  },
  "months": 24
}

The result lists each month from an empty start with its items (quantity, unit price, and amount) and totalMicroEur, the profile with defaults filled in as the assumptions, and steadyFromMonth, the first month from which every month costs the same. Here it is 14: from then on, a year of readings is kept. Assistant use is given as tokens per level. The input and output are in the API procedure reference.

Not charged

• VPN connection time.
• Overvis’s own work: the Free assistant level, documentation search and announcements.
• Usage outside a workspace.

Closed beta

During the closed beta, usage is charged to the balance so you can see what it costs, but balances cannot be topped up and a negative balance has no consequences: no workspace is suspended and no low-balance alert is sent. Payment is planned to start in approximately November 2026, and the public release is planned for approximately December 2026. Exact dates will be announced.