Introducing Cheqq

Every business runs on work that comes back.

The invoices that need chasing. The leads that need qualifying and replying to. The receipts that need filing before the quarter closes. The report that is due again on Monday, and was due last Monday, and will be due next Monday. None of it is hard. All of it returns.

The last two years handed everyone a very good assistant for that work. You ask, it answers, and the moment passes. Nothing persists. Next week the same work comes back and you start from zero — with a better draft than you would have written alone, and a system that is no further along than it was before.

Cheqq is the other thing. You describe an operation once, and Cheqq builds the system that runs it: the tables that hold the data, the automations that act on it, the app your customers see, the report that lands on Monday. Then it steps back, and the system keeps running — without a model in the middle of every step, and without you remembering to start it.

Cheqq is public today, in early beta. It is free to start, and there is no card.

Not an assistant. A workspace.

Most operations are spread across four products that do not know about each other. The database is in one. The automations are in another. The app is in a third. The deck you present on Thursday is in a fourth, holding numbers that were true on Tuesday.

The glue between them is a person. Usually the founder.

Cheqq is one workspace with five surfaces that share a single set of records:

• Data — collections with real structure: typed fields, relations, formulas, rollups, and seven ways to look at the same records.
• Automations — deterministic workflows that act on those records on a trigger or a schedule.
• Apps — portals, forms and booking pages, live at their own address, reading and writing the same data.
• Documents — real decks, sheets and reports generated from those records.
• Agents — the thing that builds all of the above, and a specialist that keeps each one running.

Because they share one set of records, the table, the automation, the app and the report cannot drift apart. There is no sync step, because there is nothing to sync.

We call one complete system a Wapp: its collections, its workflows, its agent skills, and the instructions that tell it how your company does this particular thing. Build one by describing it, or install one of the ready-made templates and reshape it by asking.

What makes it different

Your data has structure from the first minute

You do not design a schema, run a migration, or decide what a foreign key is. You say what you keep track of, and you get typed fields, relations between them, rollups that count and sum across those relations, and formulas that update themselves.

Then there is the part a spreadsheet cannot do: give a column a job and it does it for every row. Research each company and note their funding stage. Score each lead against what we actually sell. Read the attached PDF and pull out the total. The column fills itself, row by row, and shows you the sources it used — so you can check the work instead of trusting it.

Automations that don't re-roll the dice

This is the one that matters most, and it is where Cheqq diverges hardest from every other agent product.

Most AI agents run a model on every step, every time. That has three consequences people discover in production rather than in the demo: results vary between runs, cost scales with how much you use it, and you cannot read what it is going to do tomorrow — you can only watch what it did today.

Cheqq compiles your intent once, into deterministic steps you can open and read. Six kinds of trigger, real branching, real loops. The same input produces the same result on Tuesday that it produced in March. Most day-to-day runs use no AI at all, which is why running your operation costs almost nothing: you pay for the thinking, not for the doing.

Inference re-engages exactly where a step genuinely needs judgement — drafting a reply, classifying a message, pulling a total out of a badly scanned invoice — and nowhere else.

And because automation should not mean unsupervised, the steps that could cost you money wait for a named person to say yes, and chase someone else if that person goes quiet.

Apps with permissions you can actually read

Describe a client portal, an intake form, or a booking page, and it is live at its own address in minutes, sitting on records you already own.

The part other AI app builders skip is the part that decides whether you can put it in front of a customer: an app can only touch the fields you listed, through a filter the server applies on every request. Not a rule in the client that a determined visitor can widen. You can read exactly what your app can reach — and so can whoever asks you about it.

Documents that come out of the records

Not an outline you still have to build. Real .pptx decks with charts drawn as actual chart objects, real .xlsx sheets, real .docx reports — generated from your own data, themed with your brand, and editable after the fact.

A report written from your records cannot quietly disagree with them. The numbers in the Thursday deck are the numbers in the table, because nobody copied them across.

An agent for every system you run

There is one agent you talk to. Behind it, every system you install gets its own specialist, scoped to its own data and carrying its own instructions — so the agent that runs your invoicing is not guessing about your sales pipeline.

It builds the thing, and then it gets out of the way. It comes back when something needs repair, when a step genuinely needs inference, or when you want to change how the system works.

It connects to what you already have

Gmail, Notion, Linear, Stripe, Atlassian, Airtable, PayPal, Google Workspace and dozens more connect with your own account, most of them in a single approval, and every one of their tools shows up as a step you can drop into a workflow.

Not on the list? Point Cheqq at any REST API with a key, or any MCP server by URL.

And for the sites that only work when you are logged in, the Companion extension puts the agent in your own browser — in its own tab group, asking first before it buys, pays or deletes anything, and handing the keyboard back the moment you want it.

What people are running on it

The templates are the fastest answer to "what is this actually for". Thirty of them, across six categories, each arriving as a working system — seed data worth looking at, a dashboard, and at least one automation running from the moment you install it.

The shapes people build most often:

• An inbox that files itself. Every workspace gets a free AI email address. Forward to it or publish it, and receipts, leads, invoices and orders land as rows in your own tables — with replies sent where a reply is what was needed.
• A pipeline that qualifies and chases. New leads get researched and scored on arrival; the ones worth your time get a draft reply; the ones that go quiet get chased without you noticing they went quiet.
• Books that close themselves. Capture, categorise, reconcile, chase, report — every month, unprompted.
• A portal you can actually give a client. Their records, their files, their status — and nothing else.
• Research tables that fill themselves, with citations you can open.
• The Monday report, built from the numbers rather than about them.

What early beta means

Cheqq is public, and people run real operations on it today. It is also early, and we would rather you hear that from us than discover it on a Thursday.

Here is the honest shape of it. The core is solid: collections and views, workflows, integrations, documents, templates, published apps. The surface area is wide, it is growing quickly, and the newer corners are newer. You will find rough edges. When you do, telling us is by far the fastest way to get them gone — early beta is a stage where a single report still changes what ships next week.

The pace is the point of being here now: eleven releases so far this year, three of them in the past week. Every one of them is written up in the changelog, including the parts that were fixes.

What does not change while we move quickly:

• Your systems stay readable. Everything Cheqq builds — every collection, every workflow step — you can open, read and edit by hand. Nothing is a black box you have to take on faith.
• Your data stays yours. Each workspace is isolated. Published apps are default-deny. Sensitive steps can require a named human signature. Nothing crosses a boundary you did not open.
• Your work survives your mistakes. Every turn in a conversation is a restore point, and the version you abandoned is still there if you were wrong to abandon it.

Start free

The free plan is a real one: AI credits every month, your first Wapps, live apps, and active workflows. No card.

Paid plans are flat per seat — Pro from $16, Business from $39 — and add more credits, unlimited Wapps and workflows, longer run history, and the things teams need. Because your automations run deterministically, the day-to-day barely touches your credits at all. On a paid plan you can also bring your own OpenAI, Anthropic, Google or OpenRouter key and pay your provider directly; Cheqq does not mark up inference.

The whole thing starts the same way, whichever plan you are on: describe the operation you want handled, in the words you would use to explain it to somebody you just hired.

Start free → · See pricing · Browse templates

You design it. AI builds it. It runs itself — and it stays yours.