Accounting tools for managing multiple client accounts.

There is no single tool that manages multiple client accounts efficiently, and any list that names one is selling you the layer it happens to own. A firm running more than about ten QuickBooks Online clients needs five layers: practice management to say who does what by when, document intake to get source material in without chasing it, ledger access to actually reach the books, review and close to catch what went wrong, and client reporting to send something the client will read. Layers one, four and five are crowded markets with genuinely good products in them. Layer three is the one most firms never fill, and it is the layer that decides whether the other four compound or just accumulate as subscriptions.

Numbers Game is a layer-three tool. It is a QuickBooks Online connector that holds authenticated connections to every client file a firm manages and lets an accountant run work across all of them from one Claude or ChatGPT account, with a person approving every write. This article explains the whole stack, including the parts we do not build, because choosing layer three badly is how firms end up with five tools and the same week.

Products to look at: Karbon, Financial Cents, Canopy, Jetpack Workflow, TaxDome. They differ in whether they lean toward tax practices, bookkeeping practices, or both, and in how much client-facing portal they bundle. All of them solve the same core problem.

What it fixes. Work becomes visible. Recurring jobs generate themselves, deadlines exist somewhere other than a partner’s memory, a staff member leaving does not take the client’s close checklist with them, and you can finally answer “where is the Halcyon close” without asking three people.

What it does not fix. None of these tools touch a ledger. A task card reading “categorize January for Halcyon Robotics” is a card. Somebody still opens Halcyon Robotics and does the work. Practice management makes the queue legible, which is valuable and is not the same as making the queue shorter. Firms that buy practice management expecting throughput are consistently disappointed, and then blame the tool for a problem it never claimed to solve.

Ask a firm what it uses to reach client books and the answer is usually the name of a habit rather than a product: the client switcher in QuickBooks Online Accountant. That is not a tool choice. It is the absence of one, and it is the single largest constraint on a multi-client practice, because everything else in the stack eventually has to pass through it.

A ledger access layer is something that holds authenticated connections to many client files at once, and can run the same operation across all of them with per-client scoping that you control. There are three real options in 2026.

The reason this layer matters more than its profile suggests: it is the only layer that changes the arithmetic above. Practice management makes sixteen hours of navigation visible. Intake shaves the waiting. Only ledger access removes the per-client repetition itself, because it lets one request reach many books.

Here is what to demand of anything in this category, ours included. Connections should be OAuth only, so the tool never receives a QuickBooks credential. Each authorized AI client or integration should be restricted to the specific companies you name, and that restriction should be checked on every call rather than at setup. There should be a genuine read-only role that a client-side toggle cannot escalate. Revocation should take effect in seconds, not at the end of a token lifetime. Every call should land in an audit log that you can read yourself. And nothing should write to a client’s books without a named person approving it. Those are the six properties Numbers Game is built around, and they are also a fair checklist to hold us to. Our security page states the boundaries, including the ones that are not flattering.

Products to look at: Fathom, LiveFlow, Reach Reporting and Syft Analytics, which pull QuickBooks data into managed reporting, dashboards, and in LiveFlow’s case live spreadsheet models.

What it fixes. A client does not read a QuickBooks profit and loss export. They read a document with your name on it that tells them what happened. Reporting tools make that document repeatable instead of a monthly act of will.

What it does not fix. The commentary. Every tool in this category produces excellent charts and no opinion. The sentence “gross margin fell four points because you took the Halcyon job at a bad rate” is the billable part, and it is still yours to write. Which is the argument for making the retrieval cheap: the hours you save assembling numbers are the hours available to say what they mean.

One warning about sequencing. Buying reporting before ledger access is the most common expensive mistake we see, because a reporting tool makes the last mile beautiful while the first mile still takes a week. The output improves and the calendar does not.

Numbers Game was built inside AG Accounting, a firm running more than 200 QuickBooks Online clients, and every capability runs on those clients before it reaches anyone else. That is the reason the roadmap is unromantic. It is full of things like credit memos that actually apply to open invoices, and voiding rather than deleting where a third party already holds a copy of the document.

Nicole Jordan-Dahdal, who runs the fractional CFO and M&A practice PrismEdge, put the effect of collapsing the ledger-access layer this way: “A recent opex audit across three entities would normally have taken me a few days. With Numbers Game it was done in under two hours.” The detail worth noticing is which time was saved. She did not automate her judgment. She automated retrieval, and the hours that used to go into assembling numbers now go into reading them. The full write-up is here.

If you want to see the texture rather than the summary, we publish the prompts in our Prompt Library, organized by close, reconciliation, journal entries, vendor analysis and client communication.