Reference · updated August 2026

What connects, and how well

Integration marketplaces list logos. What a buyer actually needs to know is which APIs are stable, which paginate badly, which break on version changes, and which require a Windows machine somebody has to maintain. This is that list.

Is your system listed?

Tell us what you run. We will confirm what connects and what a custom integration would cost.

1 / 3
60+ systems supportedMaintenance burden statedDifficult ones named
62systems with production connectors
the effort for QuickBooks Desktop against Online
$150–500monthly maintenance per custom connector
24hreconciliation cadence on every connector
QuickBooks OnlineQuickBooks DesktopXeroNetSuiteSage IntacctSage 100 / 300AcumaticaDynamics 365 BCOdooSAP Business One

Systems with production connectors. Anything not listed is scoped case by case after a short paid discovery.

What we found

What we have learned building these.

Patterns that hold across sixty-odd integrations, several of which contradict the assumptions buyers usually bring.

API quality is uncorrelated with vendor size

Some of the best APIs we work with belong to small companies and some of the worst belong to large ones. Xero and Stripe are excellent; several enterprise products are considerably harder.

Rate limits decide the timeline

A full historical extract from several accounting APIs takes days rather than hours, entirely because of rate limiting. It is the most common reason an integration schedule slips.

Retries are the top defect source

Duplicate transactions from a redelivered webhook are the most common problem we find in inherited integrations, and they are tedious to unwind because each duplicate is individually valid.

Desktop software is not unreachable

QuickBooks Desktop has no cloud API and is entirely workable through the SDK. It is roughly three times the effort, which is why we quote it separately rather than pretending otherwise.

Two-way sync needs an owner per field

Bidirectional writes require a designated authoritative side for every field. Where neither system is clearly authoritative, we recommend against it rather than building a conflict resolver.

Maintenance is the real cost

Building a connector is one-off. Keeping it working through API versions, schema changes, and credential rotations runs $150 to $500 a month, and it is the line most proposals omit.

The reconciliation, not the connection

Any competent engineer can call an API and store the results. What makes an integration trustworthy is a daily reconciliation — counts, control totals, and spot-checked records compared between source and destination — that is written independently of the sync code.

Independence matters. A reconciliation that shares assumptions with the process it checks will confirm those assumptions rather than test them, and the failure it was built to catch is exactly the one it will miss.

Integrations do not fail loudly. They drift, and the drift is found at close by somebody who then has to reconstruct when it started.

The systems we find genuinely difficult

QuickBooks Desktop, for the reasons above. Older on-premise ERP where the only path is a database read replica and the schema is undocumented. Industry-specific applications whose vendors treat customer data as a commercial asset — that last category is the only one we sometimes cannot solve, and it is a commercial obstacle rather than a technical one.

Marketplace settlement APIs deserve a specific mention. Amazon in particular returns settlement in a structure that requires substantial work to decompose into charges, fees, refunds, and reserves, and most integrations we inherit have simply not done it.

What good looks like

The APIs we would hold up as examples share four properties: stable identifiers that survive edits, a reliable way to ask what changed since a timestamp, bulk access that does not require pagination through everything, and version deprecation announced with real notice.

Xero, Stripe, Gusto, and Shopify meet most of that. The systems that are hardest to work with usually fail on the second property — no change feed — which forces a full comparison on every sync and makes the rate limits bind much harder.

What we will not build

An integration that cannot justify its maintenance. If a system is used monthly by two people and connecting it would cost $18,000 plus $400 a month, a scheduled export is the better answer and we will say so.

We turn this work down regularly. A neglected integration silently produces wrong numbers, which is worse than the manual process it replaced, and neglect is what happens to integrations nobody can justify maintaining.

How we measured this

This catalogue lists systems with connectors running in production for at least one customer. Nothing appears here on the basis of being technically possible or on a roadmap.

Maintenance cost figures are our actual internal cost of keeping connectors working, averaged over 2025 and 2026, rather than a price list. What we charge is stated on the services pages.

API quality assessments are our engineering team’s judgement from working with them, not a scored methodology. They are opinions held with some confidence and we would revise them if a vendor improved.

Anything not listed is scoped case by case through a paid discovery of $1,500 to $3,000, which produces a written feasibility report and a fixed quote whether or not you proceed.

Questions

Common follow-ups.

What if our system is not listed?
A paid discovery of $1,500 to $3,000 establishes feasibility and produces a fixed quote. The report is yours whether or not you proceed.
How long does a supported connector take?
Days rather than weeks for the connection. The time goes into mapping and reconciliation, which is where the value is.
Who maintains it when an API changes?
We do, as part of the subscription. Connector maintenance is not billed as change requests.
Do you write back to source systems?
Where it makes sense and where a field has a clear authoritative side. Where neither system is clearly authoritative, we recommend against it.
Which APIs are best to work with?
Xero, Stripe, Gusto, and Shopify, in our experience. Stable identifiers, a real change feed, workable bulk access, and honest deprecation notice.

Send us your stack list.

We will confirm what connects, what needs building, and what is not worth connecting at all.