What exactly spends a credit?

Three kinds of action: chats (each turn), automation runs (every execution, scheduled or manual), and ask calls. Over MCP that is exactly three tools — add_prompts, run_automation and ask. The read-only tools, and reading numbers you already collected, do not spend anything.

No. A workspace has one shared balance and adding a colleague costs nothing. Two people in the same workspace draw from the same pool, so a busy automation schedule and a busy chat habit compete for it — but nobody is kept out of the numbers by a licence.

What happens when the balance hits zero?

Credit-spending actions are refused outright rather than partially performed, so you never get a half-finished automation run. Everything read-only keeps working: the workspace opens, every metric already collected is readable, and scheduled automations stay configured and resume once you top up.

How many credits does a chat cost?

Cost scales with how much work the action does — a short lookup costs less than a long analytical question or an automation that fans out across many prompts and pages. Attensira does not publish per-action credit arithmetic at the moment, because credits and the monthly response budget are currently inconsistent in the product. Check your workspace balance rather than planning around a published figure.

Will a retried API call bill me twice?

Not if you send an idempotency key. run_automation takes an idempotency_key over MCP, sent upstream as the Idempotency-Key header, and the REST endpoint takes the header directly. The same key resolves to the first run rather than starting a second billed one.

Do credits pay for the daily measurement too?

No. Measurement is what your plan buys: prompt slots, sampling depth and retention. Credits meter agent work on top of that. The two are separate numbers doing separate jobs.

Credits

You pay for work the agent does, not for seats or dashboards

A credit is the unit Attensira uses to meter agent work — the actions where a model does something on your behalf. Chats, automation runs and ask calls draw on your balance. Reading numbers you have already collected does not, and neither does having another person in the workspace.

Get Started Book A Demo

Agent work only
Chats, automation runs and ask calls. Reading your own data is free.
Not per seat
A workspace has one shared balance. Adding a colleague costs nothing.
Three MCP tools spend
add_prompts, run_automation and ask. The read-only tools do not.
Refused, never half-done
At zero, a credit-spending action is refused outright, not partially performed.

By Karl-Gustav KallasmaaLast updated September 3, 2026Behaviour on this page is checked against docs.attensira.com

What a credit actually meters

A credit meters agent work: the actions where a language model does something on your behalf rather than simply returning data you already have. Your workspace holds a balance, agent actions draw it down, and reading what you have already collected does not touch it.

Three kinds of action spend credits, whether you start them in the app or over MCP. Chats, because each turn is agent work. Automation runs, every execution, triggered by hand or by the schedule. And ask calls, which put a question to the agent and return an answer or a running session.

Opening a dashboard, filtering a chart or exporting numbers you already have does not spend credits. If your balance is moving while you do only those things, something else in the workspace is running — check your automation schedules.

Why this is a different pricing model from a per-seat tracker

Most tools in this category charge per seat, which means you pay for the number of people who might look at a chart. That is a sensible model for a dashboard, because looking is the product. It is the wrong model for an agent, because looking is not the product — the work is.

Metering the work has three consequences buyers feel. Adding a colleague costs nothing, so the person who should see the numbers is not kept out by a licence. A workspace that reads its data heavily and runs the agent rarely stays cheap. And an expensive month is an expensive month because a lot of work was done, which is a bill you can trace to a run history rather than a headcount.

It also means monitoring and agent work are priced separately: what your plan buys is measurement — prompt slots, sampling depth and retention — while credits buy the agent's effort on top of it.

How consumption behaves

Consumption is per action, not per second and not per seat. Two people in the same workspace draw from one shared balance, so a busy automation schedule and a busy chat habit compete for the same pool.

Cost scales with how much work the action actually does. A short chat turn costs less than an automation that fans out across many prompts and pages. That is why an automation you run by hand can feel cheap while the same automation on an hourly schedule is not — and why matching a cadence to the daily sampling rhythm is a budget decision as much as a data one.

Trigger a run over MCP or REST with an idempotency key and a retry does not produce a second billed run.
Creating an automation is free. Executing it is not.
Raising a row in your Inbox costs no credits, so the agent is never discouraged from telling you it is stuck.

Which MCP tools spend, and which are free

Over the MCP server, exactly three tools cost credits: add_prompts, run_automation and ask. Everything read-only is free.

Costs credits	Free
add_prompts	search, get_analytics, get_prompt, get_page
run_automation	get_session, get_account, get_ai_traffic_install
ask	list_inbox, list_automations, get_automation

What happens as the balance falls, and at zero

Nothing degrades quietly. The balance is shown in the workspace and the agent surfaces stay fully functional right up to the point where a request cannot be covered. As headroom shrinks, long-running agent work is the first thing at risk, because an automation that fans out widely needs more room than a single chat turn.

At zero, credit-spending actions stop. Starting a chat, running an automation, calling ask or adding prompts through the agent is refused rather than partially performed — you are told the workspace is out of credits, and you do not get a half-finished automation run.

Everything that does not consume credits keeps working. You can open the workspace, read every metric already collected, filter, compare, and use the read-only MCP tools. Scheduled automations do not disappear: they stay configured and resume once the balance is restored.

Where a month's credits actually go

Two workspaces on the same plan can spend very differently, and the difference is entirely visible in what they asked the agent to do.

Free → opening the dashboard, filtering a chart, exporting a CSV

Free → list_inbox, get_analytics, get_prompt over MCP

Free → creating an automation; raising an Inbox row

Spends credits → a chat turn (more for a long analytical one)

Spends credits → every automation run, scheduled or manual

Spends credits → ask, and add_prompts over MCP

Not billed twice → the same idempotency_key sent again

The lever you control is cadence. An hourly automation over data that is sampled at most once per calendar day re-reads the same measurements and bills for each pass.

What this page will not tell you

Attensira publishes what a credit is and what spends one. It deliberately does not publish credit arithmetic, and the reason is a known open issue rather than a marketing decision.

No credits-per-plan table and no cost-per-action figure. Credits and the monthly response budget are currently inconsistent in the product — a known bug we are fixing — so any arithmetic printed here would go stale or mislead.
No promise that monitoring is free of credits. Check your account before planning around a specific balance.
Credits do not buy measurement. Prompt slots, sampling depth and retention come from your plan, not from your balance.
There is no free tier and no trial. Attensira is Starter, Growth or Business.
A credit is not a seat, a query or a page. It meters agent work, and the same action costs more when it does more.

Which plans include it

Included on every plan — Starter $129/mo, Growth $349/mo, Business $699/mo.

Every plan includes a monthly credit allowance for agent work, and every plan can top up. Because credits and the response budget are currently inconsistent in the product, this page publishes no per-plan credit figures — check your workspace, and see the pricing page for what each plan includes.

Plan facts on this page are read from the same ladder that prices /pricing, so they cannot drift from what you are actually charged.

Read further

Documentation

Credits — what spends one, and what a workspace sees at zero

Documentation

Plans — prompt slots as a stock, the response budget as a flow

Pricing

Starter, Growth and Business — no free tier, no trial, no Pro plan.

Automations

Cadence is the main lever on what a workspace spends.

The agent

Each turn is metered agent work; reading your own numbers is not.

Inbox

Raising a row costs nothing, by design.

FAQ

Questions people actually ask

Answers here match the product documentation. Where the product cannot do something, this says so.

Still deciding? Compare plans or talk to us.

You pay for work the agent does, not for seats or dashboards

Agent work only

Not per seat

Three MCP tools spend

Refused, never half-done

What a credit actually meters

Why this is a different pricing model from a per-seat tracker

How consumption behaves

Which MCP tools spend, and which are free

What happens as the balance falls, and at zero

Where a month's credits actually go

What this page will not tell you

Which plans include it

Read further

Questions people actually ask

What exactly spends a credit?

Do I pay per user?

What happens when the balance hits zero?

How many credits does a chat cost?

Will a retried API call bill me twice?

Do credits pay for the daily measurement too?

What a credit actually meters

Why this is a different pricing model from a per-seat tracker

How consumption behaves

Which MCP tools spend, and which are free

What happens as the balance falls, and at zero

Where a month's credits actually go

What this page will not tell you

Which plans include it

Read further

Questions people actually ask

What exactly spends a credit?

Do I pay per user?

What happens when the balance hits zero?

How many credits does a chat cost?

Will a retried API call bill me twice?

Do credits pay for the daily measurement too?