Plans

What you pay is two things: a plan fee for the platform, and a monthly budget for the work the agents do.

• The plan pays for the platform: monitoring and analytics. It comes with a monthly allowance of credits, and it sets how many prompts you can track, how deeply each one is sampled, how long history is kept, and which features are on.
• The budget pays for the work: articles, page fixes, outreach and the rest. It is shown as the work it buys (”≈ 4 articles · 12 page fixes · 20 outreach pitches · daily monitoring” for $1,500), charged from your balance as the work is done, and it is the most that work can be charged in a month. Work we don’t deliver isn’t charged. See Your monthly budget.

So what a month costs is the plan’s fee plus the top-ups the work needed beyond the credits the plan includes, and the budget caps what the work can be charged.

​

Which plan

Signing up in the dashboard starts on Starter, $129 a month, and asks for a monthly budget beside it. You can change it any time under Settings → Billing. Starter comes with a 7-day free trial: checkout takes a card and states the charge before it takes it, and the plan fee is not charged until the trial ends unless you cancel first. Only Starter has a trial, and it is given once. Growth ($349 a month) and **Business** ($699 a month) are larger monitoring allowances: more prompts, three samples per reading, Gemini, longer history, writing your own skills, and AI traffic. The business profile, and checking every draft against your rules, are on every plan. They do not buy more work; the budget does. In the dashboard they are offered under Settings → Billing once you are set up, not at signup. Setting up from your own assistant, the plans start_workspace lists are every plan checkout sells. Past Business, Enterprise is a conversation: custom limits and terms, through Talk to sales. Paying yearly costs ten months’ price for twelve. There is no Pro plan. If you have seen one named anywhere, it is not ours.

​

What each plan includes

Models are gated by plan. Starter reads four surfaces — chatgpt, perplexity, claude and google_ai; Growth, Business and Agency add gemini. Turning Gemini on in a Starter workspace does not make it run: the scheduler queues only the engines the plan includes, so the reading never happens and the metric comes back untracked rather than as a zero. Which of a plan’s models a workspace actually reads is still yours to choose. At signup you pick where you want to be found, and when your plan starts the workspace reads the assistants you left on that the plan includes. With the signup defaults on Starter, that is all four Starter engines — chatgpt, perplexity, claude and google_ai. On Growth and above, Gemini is not in the signup defaults: a new workspace starts on the same four, and you switch Gemini on in settings. See Models for which surfaces have readers.

​

Agency

The Agency plan is no longer offered in the dashboard. Accounts already on it keep it, at $2,000 per month** or **$20,000 per year (two months free), with everything below. An agency starting now talks to sales.

• 750 prompt slots.
• 337,500 responses a month.
• Sampling depth n=3, on all five engines including Gemini.
• 50,000 credits a month, with the same rollover as the other plans.
• Unlimited history, Knowledge & Skills, and Crawler Logs.
• White label: your name, logo and colour on client reports and share links, with Attensira removed. Only Agency has it; on lower plans the setting shows an upgrade prompt that names Agency. See Branding and white label.

Unlimited client workspaces. No plan counts workspaces. An Agency account’s prompts, responses and credits are shared across every client workspace in the org. The pages built for agencies are Clients, the client reviewer link and client reports. For more, see attensira.com/solutions/agencies.

​

Your monthly budget

The budget is a monthly ceiling on what work may be charged, not a prepayment. Nothing is charged up front: work is metered as it runs, and work that never ran is never charged. Once the month’s spend reaches the budget, new work stops until you raise it or the month turns; work already running finishes. See Costs for exactly what stops. Set it under Settings → Billing, or when you sign up.

• Range. From $500 to$5,000 a month, in steps of $100.
• The month. A UTC calendar month, the same month billing counts.
• What it counts. Everything work is charged in the month: agent work, and the data reads monitoring makes. The plan’s fee is not part of it, and neither is ad spend.
• Who can change it. An organisation admin, in the dashboard. No agent, API key or MCP connection can set or raise a budget; an agent that needs more room has to ask you.
• At the budget. By default new work pauses at the budget. You can instead allow 10% over, so new work stops at 110% of it. See Pausing at the budget.

This is not the monthly response budget, which is the plan’s allowance of AI answers read for you.

​

How the budget is funded

Setting a budget turns on auto top-up and sets its monthly cap to the budget (110% of it if you allow 10% over), in the same change. Your balance is then refilled from your card as work is done, and never by more than that cap in a month. Before you save, the confirmation names the most a month can charge: “Up to $X/month will be charged to your card as work is done”.

• A top-up amount and threshold you already chose are kept. If you never chose them, the smallest top-up (5,000 credits, $50 at the one-cent list price) is used, bought when the balance falls to$10.
• Removing the budget leaves auto top-up as it is. Turning off a ceiling is not a request to stop paying for work.
• With no card on file the budget is still saved, and every workspace gets an Inbox row, Add a card so your budget can fund work. Until a card is added, work stops when the balance runs out, before the budget does. The row closes itself once a card is on file.

​

The top-up rule, and your yes to it

The budget is paid by auto top-up, so the plan step states the rule in one sentence with three numbers in it, for example:

We’ll top up $50 when your balance falls below$10, never more than $1,500 a month.

$50 is the smallest top-up we sell: 5,000 credits at the one-cent list price. Bought in one purchase, 5,000 credits fall in the 0.9¢ volume band, so the card is charged$45 (see Credits). $10 is a fifth of the$50, and the last number is the cap: the budget, or 110% of it if you chose to let work run up to 10% over. The button reads Agree and choose Starter. Pressing it saves your yes, before checkout asks for a card: the budget, whether work pauses at it or may run 10% over, the sentence’s three numbers, who agreed and when. If that save fails, the step says so and nothing is charged: try again, or skip and set a budget later in Settings. Nothing is set up by the page you land on after checkout. When your plan starts, the rule you agreed to is set once, on our side, and the budget is set as you chose it. A $1,500 budget with 10$1,500 budget that new work may run to $1,650, and auto top-up never charges more than$1,650 in the month. If you set a different budget in Settings in the meantime, yours stands. If the rule cannot be set, you get one blocked row in your Inbox, “Your monthly budget rule wasn’t set”, saying where to set it.

​

What a budget buys

A budget is turned into a volume of work at these list prices: The first $100 of every budget is set aside for monitoring and data: readings beyond your plan's own, page reads, crawls and search data. The rest is spread in a fixed mix of 4 articles to 12 page fixes to 20 outreach pitches. Every count rounds down, so the line never promises a deliverable the budget does not fully pay for. The plan step at signup starts at$1,500: Listing fixes have a list price but no share of that mix. An agent can still propose one, at its list price. These prices turn a budget into the volume we commit to, and value the credit when we fall short. They are not a price per item: what a month is charged is still metered by the work actually done, so one article can be charged more or less than $150.

​

If we deliver less than the budget committed

Work we don’t deliver isn’t charged. The budget commits us to the volume in the table above for the month, and on the 1st we check what was delivered against it. What is committed. The month’s volume, for the kinds of work we can count: articles, page fixes and outreach pitches. A budget set or changed during the month commits a share prorated by the days it was in force, so a budget set with a week left commits about a week’s work. Nothing is committed for listing fixes, because we cannot count them yet. What counts as delivered. An article or a page fix when it is completed, an outreach pitch when it is sent. Work finished on our side and waiting on you to put live (an article to paste, a Notion page to publish, a pull request to merge, a pitch drafted into your own mailbox) counts as delivered. It is never a shortfall. The credit. For each kind of work, the shortfall is what was committed less what was delivered or is waiting on you. The credit is the smaller of two amounts:

1. the shortfall, valued at the list price in force when that work was committed;
2. what you spent in the month, less the list-price value of everything delivered or waiting on you.

In plain words: undelivered work is credited back at its list price, but never more than you paid beyond the value of what you got. A budget you did not use up is not credited, because it was never charged. How it reaches you. The month is closed at 00:15 UTC on the 1st. The credit goes back on your balance, not to your card, as one credit for the month. You get an email and an Inbox notice, We credited $X back for September. A month with no shortfall closes quietly. The credit is shown against the month it is for, as Not charged on Costs, and never lowers the spend of the month it lands in. A month is credited at most once. If closing a month fails, it is tried again daily until it succeeds, and it shows in loop health as month_close_failed until then. An organisation that cancelled is still closed, and its credit lands on its balance.

​

What a budget does not buy

A budget pays for work, not for results. We do not promise any ranking, citation, mention or change in AI answers. What the work did is on your receipt, which says plainly when nothing is proven yet.

​

Prompt slots are a stock, not a flow

A prompt slot is a place in your workspace where one tracked prompt lives. Slots are a stock: the count goes down when you add a prompt and back up when you remove one. Nothing about slots resets at the start of a month, and unused slots do not accumulate. So on Growth, 150 slots means “at most 150 prompts tracked at any one moment”, not “150 prompts per month”. If you fill all 150 and want a new prompt, you remove one first, or move up a plan. Removing a prompt frees its slot immediately.

​

The monthly response budget is a flow

The monthly response budget is the other quota, and it works the opposite way. It is a flow: an allowance that is consumed as readings are collected during the month, and that resets when the billing period rolls over. Only answers that came back count against it: a read that times out or fails is given back to the budget. Spending it does not reduce your prompt slots, and freeing a slot does not refund budget you already used. These are two independent numbers measuring two different things. Slots limit how many prompts exist. The response budget limits how much answering work happens across them in a month. A workspace can be well inside its slot count and still be spending its budget quickly, because depth and model count multiply: each prompt is read on each tracked model, and each of those readings is sampled at your plan’s depth. Both are in the usage chip in the top bar, next to your credit balance: open it for Responses this month and Prompt slots, each against your plan’s limit, with Upgrade beside them. The chip turns amber when either is used up. See Home and the dashboard.

​

How sampling depth changes what you see

Sampling depth is the number of times the same prompt is put to the same model in a single reading. Starter samples once (n=1). Growth, Business and Agency sample three times (n=3), issued concurrently. Depth is a measurement decision, not a cosmetic one. Models are not deterministic, so a single answer is a single draw. At n=1, most movements in your numbers will not clear the 95% significance gate and will be reported as no proven change rather than as a rise or a fall. At n=3 you get three draws per reading, so real movements clear that gate sooner. Read Limitations before you act on a small sample. It explains exactly what the gate does and what a low n can and cannot tell you.

​

Changing plan

Change plan under Settings → Billing. Nothing changes on the click: a confirmation names the new price first. Upgrading raises your slot count and, on Growth and above, switches on writing custom skills and Crawler Logs; on Agency it also switches on white label. The business profile, and the agent reading your rules, are on every plan. Downgrading lowers your slot count and shortens retention, so plan for both: if you drop below the number of prompts you currently track, you choose which prompts to keep, and history older than the new retention window stops being available. Billing, invoices, plan changes and the budget live in Settings → Billing. For what the month has cost and where it went, see Costs. For what a credit is and which actions spend one, see Credits.