Contents
- Fast answer
- Quick facts
- What this covers
- Why it matters
- Where to go next
- Outcomes
- Step by step
- The methodology: how the match percentage is calculated
- Why connector counts are a weak signal, and what to check instead
- A worked example: a dental practice checking its stack
- The four ways an integration fails after you have bought it
- Why integration checks matter: the sprawl data
- Connector coverage by platform: the current numbers
- What changed in 2025-2026: agents talk to tools differently
- The limits: twenty tools and no depth model
- When not to use this tool, and where to go instead
- At a glance
- Mistakes to avoid
- FAQ
Fast answer
The integration checker holds twenty common business tools and fifteen agent or automation platforms. You tick the systems you run, and each platform is scored as the share of your ticked tools that appear in its recorded connector list, then sorted by that percentage. It is a coverage ratio, not a capability assessment: it is unweighted, so it does not know which of your systems is the system of record, and it has no depth model, so being listed against a tool means a connection exists rather than that it can perform the action your workflow needs. The general workflow platforms are recorded against nearly all twenty tools and will top almost any selection, and Autonoly is recorded against all of them. Use the result to rule candidates out, then verify the two or three specific actions your workflow needs in each platform's own documentation before you buy.
Facts last verified 2026-07
Quick facts
- Avg SaaS apps per organization
- 106 (BetterCloud)
- Mid-market app estimates
- 250 to 275 in some analyses
- Apps operating as shadow IT
- Roughly one third
- Zapier connectors
- 8,000+
- Make.com supported apps
- About 2,000
- n8n nodes
- About 1,500 incl. community
On this page
What this page covers
A tool visitor should leave with a decision, not just a number: build now, prepare first, choose another workflow, or follow a course path.
- 01Your tools
- 02Compatibility
- 03Gaps
- 04Recommendations
Why does this matter now?
The commonest expensive mistake in a first automation build is subscribing to a platform and discovering afterwards that it cannot write back into the one system that matters. Peripheral tools are supported nearly everywhere; the system of record is the constraint. Checking that one connection before you pay is the cheapest diligence available.
Internal path
Where to go next from this page
These links are part of the A8gent learning and conversion path. Use them to move from concept, to diagnosis, to workflow build, to course.
What you should be able to do after this
- Stack compatibility
- Integration mapping
- Gap identification
- Platform recommendations
How do you do it, step by step?
1. List only the systems the workflow touches
Not your whole stack. A first build usually touches three or four systems, and the rest are noise that makes every platform look equally good.
2. Name the specific action per system
Read new records, update this field, create this object, receive this webhook. A logo on an integrations page does not tell you whether the action you need exists.
3. Identify the system of record
Whichever system the business would not run without. If a platform cannot write back into it, the automation ends at a notification and someone retypes the result, which is most of the work.
4. Check the tier, not just the platform
Advanced connectors for heavier business systems are often gated behind higher plans, and the difference between the entry tier and the tier that includes your integration can exceed the base subscription.
5. Plan the fallback
Any platform with an HTTP request node and webhooks can reach a system with a published API. That works, and it turns a no-code build into a partly technical one, so decide who maintains it before you rely on it.
The methodology: how the match percentage is calculated
The checker holds twenty selectable business tools and fifteen platforms, each with a list of the tools it is recorded as connecting to. You tick the systems you run. For every platform, the checker counts how many of your ticked tools appear in its list, divides by the number you ticked, and shows that as a percentage. Platforms matching none of your tools are dropped. The rest sort by percentage, then by raw match count, then by the order they appear in the file.
That is a coverage ratio and nothing more. It is not weighted, so a platform that connects to your CRM but not your inbox scores the same as one with the reverse, and it has no notion of which of your systems is the system of record. It does not measure depth either: being listed against a tool says a connection exists, not that it can perform the specific action your workflow needs. Reading a contact and updating a deal stage are very different capabilities and this model cannot tell them apart.
The general workflow platforms are recorded against nineteen or twenty of the twenty tools, so they will top almost any selection you make. That is broadly true to life - the mainstream automation layers really do have the widest coverage - but it means the ranking rarely discriminates between them, and you should not read a 100 percent as a recommendation. It means every logo you ticked appears on that platform's list.
One thing to know before you read the top of the list: Autonoly is recorded against all twenty tools, so it will always show 100 percent. The Autonoly review on this site is more conservative than the widget's blurb about what that means: checked on 6 August 2026, the vendor site states three different integration figures in three places, and 40 or more named connectors is the only claim we can evidence by counting the page. Where the widget's summary line says more than that, trust the platform page.
Why connector counts are a weak signal, and what to check instead
Published connector totals are not comparable between vendors because nobody defines the unit the same way. One platform counts apps, another counts nodes including generic utilities like an HTTP request, and a third counts every trigger and action separately. Figures also move quickly and are quoted inconsistently even by the vendors themselves. Our platform research, priced and checked on 6 August 2026, records Make.com at 3,000 or more app connectors and describes Zapier as having by far the largest catalogue in the category, while declining to repeat totals we could not verify. That refusal is deliberate: we have already had to strip numbers that could not be evidenced.
What matters instead is a five-line list. Write down the systems your workflow must touch, and beside each write the specific action needed: read new records, update this field, create this object, receive this webhook. Then confirm each line on the platform's own documentation. This takes an afternoon and it settles the question that connector counts only gesture at.
Also check the tier. Advanced connectors for the heavier business systems are frequently gated behind higher plans, so the price that matters is the plan that includes your must-have integration, not the entry plan on the pricing page. The gap between those two is often larger than the base subscription.
And remember the escape hatch. Any platform with an HTTP request node and webhook support can reach a system with a published API even without a named connector. That converts a no-code build into a partly-technical one, which changes who can maintain it, but it means a missing logo is rarely fatal. Self-hostable platforms in particular are chosen for exactly this flexibility.
A worked example: a dental practice checking its stack
A dental practice ticks Gmail, Google Sheets, Calendly, Slack and Stripe and gets a long list of platforms at or near 100 percent. Useful, and also beside the point, because the system that actually decides this build is not on the list of twenty at all: it is the practice management software holding the appointment book and the recall list.
That is the general shape of the problem for any vertical business. The horizontal tools - inbox, spreadsheet, calendar, chat, payments - are supported nearly everywhere. The vertical system of record is supported nowhere by default, and it is the one the workflow depends on. Our patient recall build guide covers what connecting to a practice management system through its API actually involves, and our platform research notes that vertical systems like the ones used in dentistry, field service and fitness are typically reached through browser automation or a direct API call rather than a native connector.
So the practice's real checklist is three lines, not five. Can the platform read the recall list or a scheduled export of it. Can it write an appointment or a task back, rather than emailing the front desk to do it manually. Can it record who has already been contacted, so the same patient is not messaged twice. Everything else on the tick list is a solved problem.
The checker still earned its two minutes: it confirmed that the surrounding tools are not a constraint, which narrows the decision to one question about one system. That is a legitimate outcome. It is just a smaller outcome than a list of platforms at 100 percent implies.
The four ways an integration fails after you have bought it
The first is depth. The connector exists, it reads records, and it cannot write the one field your workflow needs to update. This is the commonest expensive surprise and it is entirely avoidable by checking the action rather than the logo. It is also why a workflow that ends in a notification instead of a record update saves so much less than the business case assumed: somebody still retypes the result.
The second is authentication. A connector held together by one person's personal account breaks when they leave, change their password or lose access to the two-factor device, and it breaks silently at the moment nobody is looking. Connect through a service account or a shared workspace identity from day one, and write down which account owns each connection. Reconnecting an integration is a five-minute job; discovering three weeks later that nothing has run since is not.
The third is rate limiting and polling. A platform that polls for new records every fifteen minutes is not real time, and a bulk import will hit a documented request limit on its first run. Zapier publishes its limits, which is more than most vendors manage, and the numbers are worth designing around rather than discovering. Where a webhook is available, use it: push beats poll on latency, on cost and on the amount of quota you burn discovering nothing has changed.
The fourth is duplication. Two channels carrying the same request, a retry after a timeout, or a workflow re-run after a fix will all produce the same action twice unless something in the design prevents it. Duplicate protection is not a connector feature, it is a design decision, and the tell is a customer receiving two identical messages four seconds apart. Decide early which field makes a request unique and check it before acting.
Why integration checks matter: the sprawl data
The average company runs far more software than its owners think, which is why checking compatibility before buying an agent platform saves real money. BetterCloud's State of SaaS research puts the average organization at 106 SaaS applications, and mid-market estimates from SaaS management vendors run considerably higher, in the 250 to 275 range for companies that count every departmental subscription. Every app in that count is a potential integration an agent platform may or may not support.
Sprawl research also explains why integration audits surface surprises. Analyses of SaaS portfolios consistently find that roughly a third of apps operate as shadow IT deployed without IT approval, and that companies carry duplicate subscriptions and orphaned apps nobody owns. Running your stack through a checker often doubles as a cleanup exercise: teams discover they are paying for two scheduling tools and an abandoned CRM before they ever get to the agent question.
Notably, the trend reversed in 2024 and 2025: organizations now use about 5 percent fewer apps year over year as consolidation replaces accumulation. AI agents accelerate this, because one agent platform with strong integrations can replace several point tools. The practical takeaway is to check integrations against the stack you intend to keep, not the stack you currently have, and use the exercise to decide which tools earn their place.
Connector coverage by platform: the current numbers
Coverage differs by an order of magnitude across the major automation layers. Zapier maintains the largest catalog, listing over 10,000 apps. Make lists more than 3,500 apps with deeper multi-step scenario control. n8n lists more than 2,000 integrations counting official and community-built nodes, though some nodes are utilities like HTTP requests rather than app-specific connectors. Those HTTP and webhook nodes matter, because they mean n8n can reach effectively any tool with an API even without a named connector.
Raw connector count is a starting filter, not a verdict. A named integration can mean anything from full read-write access across all objects to a single trigger. The checker's depth ratings exist because the gap is enormous in practice: a support agent that can read a HubSpot contact but not update deal stages automates half a workflow, and the unautomated half is usually the annoying half. Before committing, verify the two or three specific actions your workflow needs, not just the logo on an integrations page.
Dedicated agent platforms flip the tradeoff. Tools like Gorgias and Intercom offer far fewer total integrations than Zapier, but their connections to Shopify or to a help desk stack are deep, purpose-built, and maintained as core product. The reliable pattern for small teams: use a vertical tool where its deep integrations cover your core workflow, and bridge everything else through Zapier, Make.com, or n8n rather than forcing one platform to do both jobs.
What changed in 2025-2026: agents talk to tools differently
The biggest structural change is the rise of standardized agent-to-tool protocols, led by the Model Context Protocol, which lets AI agents discover and call tools through a common interface instead of bespoke per-app connectors. Major platforms added support across 2025, and the effect is that the question is shifting from does platform X have a connector for tool Y toward does tool Y expose its capabilities in an agent-readable way. Expect integration checkers, including this one, to track protocol support alongside classic connectors.
The second change is that integration became the pricing battleground. As per-task model costs fell, platforms increasingly differentiate on how many systems an agent can act in, gating advanced connectors like Salesforce or NetSuite behind higher tiers. When comparing platforms, price the tier that includes your must-have integrations, not the entry tier, because the delta is often larger than the base subscription itself.
Third, middleware stopped being a fallback and became the norm for agent deployments. Because agents need to read from and write to many systems at once, a hub-and-spoke design with n8n or Make.com in the middle is now the standard architecture for small-business agents, with the agent making decisions and the workflow layer executing them across apps. This design also contains failures: when one connector breaks, you fix one spoke instead of rebuilding the agent.
The limits: twenty tools and no depth model
Twenty selectable tools cannot represent a real stack. There is no practice management system, no field service platform, no property management system, no legal case management, no accounting ledger beyond one option and no phone system on the list, and those are precisely the systems that decide vertical builds. If your system of record is missing, the result tells you about your peripheral tools only.
There is no depth rating in this tool. Any claim that a result distinguishes full read-write access from trigger-only access would be false, and we have corrected our own copy where it said otherwise. Depth is real and it matters enormously, but you establish it from the platform's documentation for the specific actions you need, not from this widget.
There is no freshness guarantee on the underlying lists either. Connector catalogues change monthly, and a platform that did not support your tool last quarter may support it now, or may have moved it to a higher tier. Verify on the vendor's own integration directory before you buy, and prefer a platform page that carries the date it was checked.
Finally, coverage is not the same as fit. A platform can connect to every tool you own and still be the wrong choice because its billing unit punishes your workflow shape, because it cannot hold state across a multi-day approval, or because nobody on your team will maintain it. Coverage is a filter that removes candidates. It never picks the winner.
When not to use this tool, and where to go instead
Do not use it if your system of record is a vertical product. Read the industry page for your sector, which names the systems we actually connect to in that field, and the relevant build guide, which shows what the connection involves in practice.
Do not use it as a price comparison. Coverage says nothing about cost, and the cost difference between billing units at your volume is usually the larger number. The platform pages carry dated pricing and the billing unit for each candidate.
Do not use it before the workflow exists. A tick list of everything you own produces a comparison of everything, and most of what you own is irrelevant to the first build. Specify the workflow, list only the systems it touches, then check those.
And if the honest answer is that your systems do not connect cleanly to anything, that is worth knowing early rather than after a subscription. It usually means the build needs an API integration rather than a connector, which is a different project with a different price. Our build cost page covers what that looks like, and the free automation audit will tell you which of the two you are actually facing.
At a glance
Integration coverage across major automation layers (2026)
| Platform | Connector count | Model | Best fit |
|---|---|---|---|
| Zapier | 10,000+ apps | Hosted, trigger-action | Broadest coverage, non-technical teams |
| Make.com | 3,500+ apps | Hosted, visual scenarios | Multi-step logic at lower cost |
| n8n | 2,000+ integrations | Open source, self-hostable | Custom APIs, data control, agent hubs |
| Vertical agent tools (Intercom, Gorgias) | Dozens, deep | Native product integrations | Core workflow depth over breadth |
What mistakes should you avoid?
- Treating a 100 percent match as a recommendation. It means every logo you ticked appears on that platform's list.
- Comparing published connector counts, which no two vendors define the same way.
- Assuming a listed integration supports the specific action your workflow needs.
- Ticking your whole stack instead of the three or four systems the first build touches.
- Forgetting the vertical system of record, which is usually the only hard constraint.
FAQ
- How does the integration checker calculate a match percentage?
- It counts how many of the tools you ticked appear in a platform's recorded connector list, divides by the number you ticked, and shows the result as a percentage. Platforms matching none of your tools are removed, and the rest sort by percentage, then by raw match count. It is an unweighted coverage ratio, so it does not know which of your systems is the system of record, and it treats a connection to your calendar as equal in importance to a connection to your CRM. Use it to eliminate candidates, not to choose one.
- Does a listed integration mean the platform can do what I need?
- No, and this is the gap that costs people money. A listed integration can mean full read and write access across every object, or a single trigger that fires when a record is created. There is no depth rating in this tool, and any copy suggesting otherwise was wrong. Write down the specific actions your workflow needs - read new records, update this field, create this object - and confirm each one in the platform's own documentation before you subscribe. That check takes an afternoon and it is the one that matters.
- Why do the big automation platforms always come out on top?
- Because they are recorded against nineteen or twenty of the twenty tools in the list, so any selection you make will be almost fully covered. That reflects reality - the mainstream workflow layers really do have the broadest catalogues - but it means the ranking rarely separates them from each other. When several platforms tie near 100 percent, the decision moves to the things this tool does not measure: billing unit at your volume, how the platform handles a multi-day approval, and who on your team will maintain it.
- My main system is not in the list. What now?
- Then the checker has told you about your peripheral tools only, which is still worth two minutes but is not the answer. Vertical systems - practice management, field service, property management, case management - are the usual constraint, and they are typically reached through a published API, a scheduled export, or browser automation rather than a named connector. Read the industry page for your sector and the relevant build guide, both of which describe what connecting to those systems actually involves and what it changes about the build.
- Are connector counts a good way to compare platforms?
- They are a weak signal, because no two vendors count the same way. One counts apps, another counts nodes including generic utilities, a third counts every trigger and action separately, and the totals move constantly. Our platform reviews record Make.com at 3,000 or more app connectors and describes Zapier's catalogue as the largest in the category, checked on 6 August 2026, and deliberately does not repeat totals we could not evidence. Compare on the five systems you actually need, and ignore the headline number.
- Is a native integration better than connecting through a middleware platform?
- Usually, but not always, and the difference is smaller than it sounds. Native connections are faster, have fewer failure points and often expose more of the underlying API. A middleware connection that works today beats a native connection you are waiting for, and a hub-and-spoke design where one workflow platform sits in the middle has a real advantage: when one connector breaks you fix one spoke instead of rebuilding the agent. Start with whatever connects your systems now and revisit when volume or reliability makes it worth it.
- How often should I re-check integration coverage?
- Before any purchase, and again whenever you add a system. Connector catalogues change monthly, connectors get moved between pricing tiers, and vendors are acquired. The safest habit is to open the vendor's own integration directory on the day you decide, confirm the specific actions rather than the logo, and note the date. Any comparison that does not carry the date it was checked - including this one - should be treated as indicative and verified before money moves.
Sources & further reading
Was this page helpful?
