Skip to content

QuickBooks

Integration · Payments and accounting

Connect an AI agent to QuickBooks Online.

An AI agent can create and read customers, invoices, bills, payments and reports in QuickBooks Online through Intuit’s Accounting API. It suits invoicing from other systems, bill capture and debtor follow-up. Plan for Intuit’s app review for production access and its partner programme, which meters some API usage.

What can an agent do in QuickBooks Online?

Intuit’s QuickBooks Online Accounting API exposes the main accounting records: customers and suppliers, products and services, invoices, estimates, bills, payments, accounts and journal entries, plus standard reports. Typical agent jobs are:

  • Invoice from the work system. When a job, session or delivery is marked complete elsewhere, create the matching invoice with the right items and customer.
  • Estimates to invoices. Turn an accepted estimate into an invoice, or prepare estimates from a conversation. See quote generation.
  • Bill capture. Read supplier invoices from an inbox and create draft bills with the original attached. See invoice processing.
  • Debtor follow-up. Check open invoices and draft reminders for a person to send, adjusted to each customer’s history.

What does it take to connect?

Access goes through an app on Intuit’s developer platform, using OAuth 2.0. A QuickBooks admin authorises the app for your company file, choosing which company it can reach.

  1. Build and test against a QuickBooks sandbox company using development keys.
  2. Complete Intuit’s production requirements, which include an app assessment, before using production keys against real books.
  3. Decide which Intuit partner tier the app belongs in, and who owns the developer account.
  4. Authorise the production connection with a QuickBooks admin, and store refresh tokens securely.

Intuit charges apps through its App Partner Program. Some read-heavy API calls are metered, with a monthly allowance on the free tier and fees above it on paid tiers. These fees fall on the app owner, so agree up front whether that is you or your build partner, and design the integration to avoid needless reads.

What are the limits and gotchas?

  • Rate limits apply per company and per app. Use batch operations where the API offers them, and webhooks or change queries instead of re-reading everything.
  • Country editions differ. Tax handling and some fields vary by the country edition of QuickBooks. Test with the edition you actually use.
  • QuickBooks Desktop is different. The online API covers QuickBooks Online. Desktop versions need a separate, older integration route.
  • Sync conflicts. If staff edit an invoice while the agent updates it, one change can be lost. Have the agent check the record’s version before writing.

Before you connect QuickBooks: a checklist.

Checklist

Connecting QuickBooks safely

Native apps, connectors or a custom agent?

Choosing an approach
ApproachGood fit when
QuickBooks’ own features and app marketplaceA ready-made app already covers the job, such as receipt capture or a payment-provider sync.
[Zapier](/integrations/zapier/) or similarThe rule is fixed: a new order creates an invoice, a paid invoice updates a sheet.
A custom agentThe input is messy (emails, PDFs, WhatsApp messages), matching needs judgement, or several systems must stay in step.

If you are choosing accounting software rather than integrating with what you have, Xero and Sage are common in Ireland too. Pick the one your accountant supports best; the integration work is similar.

How do you keep the connection secure?

Request only the accounting scope the workflow needs, keep the developer account and tokens owned by your business, and log every record the agent creates or changes with a link back to the source document. QuickBooks keeps its own audit log of changes, which lets your accountant see which entries came from the integration. Review the connection when staff or suppliers change. The AI agent security guide covers token storage in more detail.

What does invoicing from completed work look like?

For example, a hypothetical training provider runs courses booked in one system and invoices in QuickBooks. An integration could:

  1. Watch for courses marked as delivered in the booking system.
  2. Find or create the customer in QuickBooks, matching on company name and email.
  3. Create an invoice with the correct item, price and purchase order reference.
  4. Send it, or leave it as a draft for review if anything did not match exactly.
  5. Record the QuickBooks invoice number back in the booking system.

Most of these steps are plain automation. The agent earns its place in the exceptions: a purchase order in an email attachment, a company that changed its name, or a customer who asked for the invoice to be split.

Measure it by days from delivery to invoice, invoices that needed correcting after sending, and the number of exceptions each week. If exceptions are rare, you may not need the AI part at all, and plain automation will be cheaper to run.

Common questions

Can the agent match payments to open invoices?

Yes, it can read incoming payments and open invoices and propose matches, recording the payment against the invoice when the match is clear. Partial payments, combined payments and missing references are where it should flag a person instead of guessing, since an incorrect match is harder to unwind than a delay.

Do we need to pay Intuit to connect our own books?

Intuit’s partner programme has a free tier with a monthly allowance of metered calls, which may be enough for one company’s workflow. Larger volumes or other requirements can move an app to a paid tier. Check Intuit’s current partner terms against the expected volume before building.

Can we test without touching our real accounts?

Yes. Intuit provides sandbox companies for development, and the whole workflow should be built and tested there first. Before going live, run it on a small, real set of transactions with drafts only, and compare its output with what your bookkeeper would have entered.

Bring us the task.

Tell us what happens today, which tools are involved and where it goes wrong. On a free 30-minute review, we’ll work out whether an agent, an automation or neither is the right fix, and what a sensible first release would be.

Start a project.

Two people on a sofa with laptops, planning a project beside a whiteboard of sticky notes

Book a call or send us a message. We read everything.

tash@neu.ie
Working with teams in Ireland and worldwide.

Get in touch

Free 30-minute AI review.