Limits and plans
Free for learning, Self-serve for a team that runs it, and Done for you.
Overscore has three plans: Free, for learning; Self-serve, the whole product for a team that runs it themselves; and Done for you, where we build Overscore around your business and run it with you. Every plan is a machine of your own, a whole processor with 2 GB of memory, that sleeps when nobody is there.
Free
Free is for learning Overscore: a machine of your own (a whole processor and 2 GB of memory) that sleeps while you are away, your own files and Google Sheets as data, up to three connected systems (your store, your books, your payments or your CRM), and your own Claude Code working through ovs and MCP. Dashboards, metrics, views and the context library all work on it, and a dashboard refreshes when you reload it. On Self-serve and Done for you, live updates are on by default and a dashboard refreshes itself when its data changes.
Its limits: 100 tables, 1,000,000 rows as you see them and 1 GB of data in all; 10 awake hours a month (only the time the machine is awake counts, from your first visit until a few minutes after your last); three connected systems, each pulled once a day. Its worker has 10 hours of background time a month (Background time, below). Overscore's AI sorts what you add to the library, on us, and counts toward none of them. So nobody runs up a bill by accident, sorting pauses until the 1st once a month's sorting reaches a ceiling of ours, and what you add then waits in the Inbox, unsorted, for you or your own Claude to sort.
What Free leaves out: a fourth connected system, webhooks, database links (Postgres, or a copied warehouse such as BigQuery) and pipelines of your own, sending rows to a system, anything else on a timer (a model or an alert runs when you run it; an email goes to you, by Send me a test), Python models (a model is SQL, saved as a view), web searches, and Overscore's own AI past sorting: the chat, the dashboard builder, the interview, quick answers, reports, the facts a document suggests, the nightly re-read and an email's AI paragraph. A new company's first three chat messages are the one exception, on us, as on Self-serve. Your own Claude does those. The app does not offer them on Free: where one would start, it says what your own Claude does, with the way to Settings, Agents, and Sources and Destinations say what Free takes.
Connected systems on Free
Free connects three of the apps on the Sources screen: Shopify, Stripe, QuickBooks, HubSpot, PostHog, Google Ads and the rest. Each is pulled when you connect it, then once a day together with your Google Sheets, at a time of your company's own between 06:00 and 10:00 UTC (each source's page shows it), so your machine wakes once for all of them and sleeps again when they are done. Those minutes count toward the 10 awake hours. You can pull one by hand once an hour, and a sheet whenever you like. A fourth is refused before you fill anything in; disconnect one to make room. A company moved to Free with more than three keeps them all, and the three connected first are pulled.
Rows count as you see them: a connected system's records count once, in its tables. The copy of everything it has ever sent, which Overscore keeps behind the scenes, is not counted in your rows; its files count toward the 1 GB, since they take real disk.
A pull is counted before any of it lands: the rows of the tables it makes, the tables, and its size. One that would pass a limit is refused whole, so nothing of it lands and nothing you have changes: its source shows the limit's own sentence, the Inbox's card of the month says it too, and the next day's pull tries again.
At a limit, what would go over is refused, with a sentence that says which limit and where you are; nothing you have is removed. Past the awake hours, only Settings answers until the 1st of the month (UTC). To go past any limit, an admin presses Use this for my company in Settings, Plan and billing, with a line about the company if they like: Kyle is in touch within a day, and moving to Self-serve or Done for you keeps your files, views, metrics and dashboards as they are.
Self-serve
Everything Overscore does, run by your own team: five connected systems from the Sources screen, each synced every hour, and any number of webhooks, database links, pipelines of your own, destinations, files and Google Sheets; models as views, tables or Python; 10 GB of data; four jobs running at once, each on a worker of its own. Its machine sleeps when nobody is there and wakes on time for its own syncs, alerts and emails, with no limit on its awake hours. Its workers have 50 hours of background time a month (Background time, below). Its AI is a limit a month we set with you, $10 to start; sorting what you add to the library is on us. A sixth connected system is refused before you fill anything in: disconnect one, or ask us. A sign-up code from us opens it.
Done for you
We run it for you. The monthly fee covers running it and a set number of build hours each month for new work. The systems to connect, the dashboards and reports to build, sync frequency, capacity and AI usage are agreed for your business before work starts: any number of systems, synced every 15 minutes, eight jobs running at once, 250 GB, its machine kept awake when we agree it should be, and AI up to a limit we set with you ($200 a month to start), on your monthly invoice. See how it works.
When your machine sleeps
Every company's machine is a whole processor of its own with 2 GB of memory, and on every plan it sleeps when nobody is there. Asleep, it costs almost nothing and keeps everything: it wakes for the next request, for its own work when that comes due (a sync, a sheet's pull, an alert, an email), and when someone of the company opens the app, which wakes it while the first screen loads. A dashboard left on a screen does not keep it awake, and wakes it only to answer the page when its data changes.
The work nobody waits on runs beside it, on your company's own workers: syncs, model builds, the small tables your browser works a dashboard's numbers out from, alerts, emails and sends to other systems. They are machines of your own too, each a whole processor with 2 GB, and they sleep until there is work and wake in under a second, so your machine stays free for the people and agents using it. Free has one, Self-serve four and Done for you eight, and that many jobs run at once. The rest wait their turn, with what someone asked for (a first sync, a model run by hand) ahead of the scheduled work. A sync or a model build that runs out of memory on a worker is tried once more on your company's own 4 GB machine, and starts there from then on. Its page says so and since when, and someone who may change it can move it back to the workers.
With workers switched on, what you ask for runs on them too: a table published, Run now, Sync now, ovs run, an agent's build or first pull and a fix pressed on a card in your Inbox each take the front of the line, after a webhook's rows, and say where they are, waiting for a worker, then building or pulling on worker-2. When a busy company needs more, we raise its number of workers with you, and that many of its jobs run at once.
Plan and billing says under the plan that your machine sleeps when nobody is here, when it last woke and how long it had slept before (the first page after a sleep takes a second or two longer), and when its own work next wakes it, such as a sheet's pull or the daily pull. On Done for you we can keep it awake all the time, and then it says Always on. On Self-serve and Done for you, one line under the plan's card says the plan, the AI limit and the machine are set by Overscore, and to ask your Overscore contact to change one; nobody in the company changes them. On Free, This month says its awake hours against its ten, and on Free and Self-serve the background time used against the plan's. ovs plan prints the same.
Background time
Your company's workers are awake only while they work, and that time is your background time. Free has 10 hours of it a month and Self-serve 50, and Done for you has no limit. Plan and billing shows how much is used in This month, and ovs plan prints it.
Past it nothing fails and nothing is lost. Until the 1st of the month (UTC) each scheduled sync runs once a day, soon after midnight UTC, and its source's page says so. The models it updates build after it, as always. What you ask for still runs at once (Sync now, a first sync, Run, a fix pressed on a card in your Inbox), and so does everything that is not a scheduled sync:
- a webhook's rows
- your emails and alerts, each on its own time
- sends to other systems, and a model on its own schedule
- the models that read a file you load or a table you fill in
On the 1st the count starts again and the syncs go back to their schedules by themselves. Moving to another plan lifts it at once.
Getting in
Joining an existing company by invitation needs no sign-up code. Use the address on your invitation. Creating a company is by code. The code you're given decides the plan your warehouse starts on: a code from the community opens Free, and the one we give a team that runs it themselves opens Self-serve. Sign in with Google or an email and password; the code is asked for each time a company is made, a second or a tenth too, since it decides that company's plan. A new company is two boxes, the code and the company's name. Its address is the name made plain (app.overscore.tech/damn-filters), checked as you type: a tick when it is free, and when it is taken, change the name. There is never a number or random letters after it. Once it is built, four doors: talk about the business (the interview, which starts with your website), connect a source, go to the app, or take the tour, first for someone with no other company. On Free the first screen says the limits once, and Settings, Plan and billing shows where you are against them. When Free is full, a new Free sign-up goes on a waiting list: nothing is built yet, and we email you when yours is ready. Self-serve and Done for you never wait.
Reading your bill
Nothing is charged by card. Done for you is invoiced by hand each month, with the AI it used; the invoice gives the agreed amount and the payment date. Free costs nothing, and on Self-serve the AI is ours up to its limit. Activity, Usage is about what people use; its Plan and billing link takes you to the plan.
Today's meters
On Plan and billing, Data downloaded is the amount served through downloads and live links today; Processing time and Query time show the time used against today's limits. Each amount stays with its unit.
This month, on Free
On Free, Plan and billing shows This month beside today's meters: tables, rows, data stored, awake hours and background time, each as used of allowed. Tables and rows are what your own tables hold now, the rows as you see them (a connected system's copies kept behind the scenes count as tables and in data stored, not in rows), and data stored is those tables and the files in Files, the originals your tables came from included (Overscore's own records and what is in the trash are not counted; deleting a file frees its room); awake hours start again on the 1st of the month (UTC), and from nought the day a company moves to Free: hours on another plan never come along. At 80 percent of any of them, the admins get one card in the Inbox for the month, however many things notice at once, which closes itself when the month turns, or when the company moves to a plan without these limits. ovs plan prints the same numbers.
Your AI limit
Home and Settings, Plan and billing show the AI used this month against your company's limit, a number of dollars a month we set with you: Self-serve starts at $10, and on Done for you it is agreed with you. Chats, building dashboards and building models count toward it. Sorting what you add to the library is on us on every plan and never counts. It has ceilings of ours so nobody runs up a bill by accident, a day's on every plan and a month's on Free and Self-serve. Past one, sorting pauses until midnight UTC or until the 1st, what you add waits in the Inbox, unsorted, for you or your own Claude to sort, and everything else keeps working. The amount is the whole company's, whoever asked. When the AI stops at the limit, the message says so, who to ask to raise it, and that the AI starts again on the 1st of next month (UTC); syncs, SQL, models, the inbox and sorting keep working. Your own Claude Code, connected in Settings, Agents, works on your own Claude plan and is never counted.
AI on its own
AI on its own is the AI that acts without being asked. It reads failed runs, looks into alerts, writes email summaries and answers a Python model's asks. It has a ceiling of its own inside your AI limit, $2 a month on Self-serve and $5 on Done for you to start, which we set with you, and Plan and billing shows its spend this month under the AI's. An admin switches it off and on in Settings, Company. Off, or at its ceiling until the 1st, the watcher picks its fix by a rule of thumb, an alert's card says what it found without its sentences on top, an email goes without its sentences on top, and a model's ask is told so, while the chat goes on. Free has none: its AI only sorts.
When you need more
If you need more capacity, faster syncs or more build hours, talk to Kyle and we'll adjust your plan.