API integration cost is the line most businesses discover late — after the software is chosen, the contract is signed, and somebody finally asks how the new system will talk to the old one. The quote arrives as a single figure with no working shown, and it is either alarmingly high or suspiciously low. Both are symptoms of the same problem: nobody has separated the work of moving data between two systems from the work of keeping it moving for the next three years.
This guide breaks that number apart. It covers what sits inside a realistic API integration cost, the UK price bands by project type, the pricing models suppliers offer and what each one quietly hides, the recurring platform fees that never appear in a build quote, and the maintenance percentage you should be setting aside from the first day rather than the first failure.
It also gives you a way to reach your own figure before you speak to anybody. Count endpoints rather than systems, score data complexity honestly, and add contingency for the parts of the vendor’s API nobody has read yet. Businesses that work through that method land within roughly twenty per cent of the eventual invoice — close enough to make a decision on, and close enough to recognise a quote that has left something out.
Table of contents
- What API integration cost actually covers
- API integration cost by project type
- The pricing models suppliers offer
- Hidden costs that wreck an API integration budget
- Platform and licence fees you pay every month
- Security, compliance and API integration cost
- Maintenance: the recurring API integration cost nobody quotes
- In-house versus outsourced API integration cost
- How to estimate your API integration cost in one afternoon
- Reducing the bill without cutting corners
- Frequently asked questions about API integration cost
- Turning the estimate into a decision
What API integration cost actually covers
Suppliers quote wildly different numbers for what sounds like the same job because they are pricing different amounts of work. A complete API integration cost has five components, and a quote missing any of them is not cheaper — it has simply moved that work onto you, or onto a later invoice with a change-request number attached.
Discovery and documentation review
Before a line of code exists, somebody reads the API documentation of both systems and confirms the integration is possible as specified. This is where you learn that the CRM exposes contacts but not contact history, or that the accounting platform’s sandbox behaves differently from production. Expect £1,500–£4,000, or five to fifteen per cent of the project. Skipping discovery does not remove the work — it relocates it into development, where it costs considerably more.
Field mapping and transformation logic
Two systems rarely agree on anything. One stores a full name, the other first and last; one uses ISO country codes, the other free text; one treats a deleted record as archived, the other removes it outright. Every disagreement becomes a rule somebody writes and tests. Mapping is the single largest driver of an API integration cost, and the only honest way to size it is to count the fields before quoting.
Error handling, retries and idempotency
The happy path is perhaps a third of the work. The rest is what happens when the target system is down, the payload is rejected, or the same message arrives twice. Handling duplicates safely — idempotency, in engineering terms — is not optional for anything touching orders, invoices or payments. Integrations quoted without it tend to fail during their first genuinely busy week.
Monitoring, logging and alerting
An integration nobody watches is an integration that has already failed silently. You need to know within minutes, not at month-end when the figures refuse to reconcile. Alerting, a readable log of every message, and a mechanism to replay failed items are what separate a professional build from a script on somebody’s laptop. Budget for them explicitly, because this is where a cheap API integration cost usually hides its omissions.
Documentation and handover
Whoever maintains this in two years will not be whoever built it. A short document naming the endpoints used, the authentication method, the mapping rules and the known failure modes costs about a day to produce and saves a rebuild later. Insist on it in the statement of work rather than requesting it at the end.
API integration cost by project type
Price bands are more useful than averages, because integration projects do not cluster around a mean. The API integration cost of any given project depends almost entirely on direction, volume and how much the two systems disagree with each other. The bands below reflect UK market rates in 2026 for a competent supplier, excluding platform licences.
Simple one-way push: £2,000–£6,000
One system sends data to another on a schedule or a trigger — a web form into a CRM, orders into a warehouse system, a nightly customer export. Few fields, no conflict resolution, and failures resolved by re-running the job. Most organisations’ first integration lives in this band, and it is genuinely quick: two to four weeks including testing and handover.
Two-way sync with conflict rules: £8,000–£20,000
Both systems can change the same record, so somebody must decide who wins. That single requirement roughly triples the API integration cost, because every field needs a precedence rule, a change-detection mechanism and a loop-prevention guard. Bidirectional sync between a CRM and an ERP sits here, and six to ten weeks is a realistic delivery window.
Multi-system orchestration: £25,000–£80,000+
Three or more systems, a shared data model, and business logic spanning all of them — order-to-cash, or the joiner-mover-leaver process across HR, identity and payroll. At this point you are not integrating, you are building a small platform, and the API integration cost should be planned like one: in phases, with a defined first release that delivers value on its own.
Real-time and high-volume: add 30–60%
Anything requiring sub-second latency, guaranteed message ordering, or throughput above a few thousand messages an hour needs queueing, back-pressure handling and a different testing regime altogether. It is rare in mid-market work and expensive when genuinely required — yet frequently specified when a five-minute delay would have been perfectly acceptable. Challenge that requirement before paying for it.
The pricing models suppliers offer
Two suppliers can quote an identical figure and expose you to completely different risk. How the API integration cost is structured matters as much as its size, because the structure decides who pays when the vendor’s API turns out to be worse than its documentation promised.
Fixed price
Fixed price works when scope is genuinely known: a documented API, an agreed field list, one direction of travel. The supplier carries delivery risk and prices it, typically at a fifteen to thirty per cent premium. It fails badly against vague scope — every discovery becomes a negotiation, and the working relationship turns adversarial somewhere around week three.
Time and materials
Honest for exploratory work and dangerous without controls. UK day rates in 2026 run roughly £450–£650 for a mid-level integration developer, £700–£950 for a specialist or solution architect, and £550–£750 blended at an agency. Cap the exposure: agree a not-to-exceed figure, a weekly burn report, and a genuine checkpoint at the halfway mark.
Fixed-price discovery, then build
This is the model that works most often. Pay a small fixed fee for discovery and a written estimate, then decide whether to proceed. It costs a fortnight and converts an API integration cost guess into something defensible — the same logic that makes a proper software discovery phase worthwhile on any larger project.
Retainer and integration-as-a-service
For businesses running several connections, a monthly retainer covering monitoring, fixes and small changes usually costs less than repeated project engagements. Expect £800–£2,500 a month depending on the size of the estate. It also solves the ownership problem, because somebody is contractually on the hook when a vendor retires an API version.
The overruns are rarely in the code. They sit in assumptions nobody tested before the estimate was signed off, and they recur with such regularity that they are worth walking through one at a time before you commit a budget.
Rate limits and pagination
A vendor API permitting sixty calls a minute will not move a hundred thousand records overnight, and discovering that in week five means redesigning around batching and back-off. Ask for the limits in writing during discovery. Rate limiting is the most common reason an integration passes testing and then collapses against real data volumes.
Authentication that is not OAuth
Modern platforms use OAuth 2.0, and that path is well trodden. Legacy systems use IP allowlists, rotating keys in custom headers, SOAP with WS-Security, or a VPN tunnel somebody must configure and then maintain indefinitely. Each variant adds days to the API integration cost, and the tunnel adds an infrastructure dependency that never goes away.
Undocumented and legacy APIs
Documentation that is out of date is worse than no documentation, because the estimate was built on it. The API integration cost for a legacy on-premise system should therefore carry a larger contingency — thirty to forty per cent rather than twenty — purely because discovery cannot be completed from the outside.
Sandbox access and test data
You cannot test an order integration against live orders. If the vendor charges for a sandbox environment, or does not offer one at all, the project needs a workaround that consumes real days. Several enterprise platforms gate sandbox access behind a partner agreement that takes weeks to obtain, which is a schedule risk long before it is a cost.
Vendor API licence tiers
A growing number of SaaS products put API access behind a higher licence tier, or meter it per call. This can dwarf the build itself: a plan upgrade at £30 per user per month across eighty users is £28,800 a year, several times the one-off API integration cost of the work. Confirm the tier and the metering before anybody scopes anything.
Change requests from the other side
The system you integrate with will be upgraded by somebody else, on their timetable. Two vendor releases a year, each requiring a day of regression testing on your side, is an entirely normal recurring commitment. It belongs in the annual budget rather than arriving as an unexpected invoice in month eight.
Platform and licence fees you pay every month
Custom code carries no licence fee but all of the maintenance. Integration platforms invert that trade, and the inversion sits at the heart of any honest API integration cost comparison. Neither option is universally cheaper — the crossover depends on message volume and on how many integrations you expect to run over three years.
Low-code iPaaS subscriptions
Tools such as Zapier, Make and Power Automate start near £15–£30 a month and are genuinely the right answer for low-volume, low-risk connections: a form to a spreadsheet, a lead into a CRM, a notification into a channel. Two people can build one in an afternoon, and that speed is real value for straightforward workflow automation.
Task-based pricing that scales badly
The trap is the meter. Per-task pricing is trivial at a thousand operations a month and painful at two million: a £29 plan becomes a £900 one as volume grows, and the eventual migration to custom code costs what you thought you had saved. Model twelve months of projected volume rather than this month’s usage.
Enterprise middleware
MuleSoft, Boomi and comparable platforms price from roughly £25,000 to well beyond £100,000 a year, and they are bought for governance, reuse and vendor support rather than for any single connection. They only justify the API integration cost across a portfolio — typically ten or more integrations sharing common assets and standards.
Hosting for custom integrations
A bespoke integration still needs somewhere to run: a small container, a queue, a secrets store and monitoring. Realistically that is £30–£200 a month on a mainstream cloud platform, which is one reason cloud adoption decisions and integration decisions belong in the same conversation rather than in separate budgets.
Security, compliance and API integration cost
Integrations move data across trust boundaries, which makes them a cybersecurity question as much as a development one. Security work is not an optional extra layered on top of the API integration cost — it is part of the cost, and leaving it out is precisely how a cheap quote becomes an incident report.
Credential and secret management
API keys living in a config file on a developer’s laptop, or in a shared spreadsheet, remain depressingly common and are indefensible. Secrets belong in a managed vault with rotation and an audit trail. That is half a day of work at build time and a disclosable breach if skipped, which makes it the cheapest insurance in the entire project.
Least privilege and scope
Integrations are routinely handed full administrative access because it is faster than working out which permissions they actually need. Scope every credential to the endpoints it calls and nothing more. When the integration is eventually compromised or simply misconfigured, that scoping is what limits the damage to something recoverable.
Data protection and lawful basis
Moving personal data between systems is processing it. You need to know which fields cross the boundary, where they come to rest, how long they are retained and who acts as processor. This is ordinary data protection work, and it must happen before the pipe is built rather than during the incident afterwards.
Testing and independent review
A penetration test against an internet-facing integration endpoint costs £1,500–£5,000 and is proportionate for anything handling payments or personal data at volume. For internal-only integrations, a code review against the OWASP API security risks is normally sufficient and adds perhaps a day to the API integration cost.
Maintenance: the recurring API integration cost nobody quotes
A build is a project; an integration is a relationship. Budget fifteen to twenty-five per cent of the build figure every year for maintenance, and treat that as the true API integration cost of ownership rather than an optional support package you might decline.
API deprecations and version changes
Vendors retire API versions on their own schedule, usually with six to twelve months’ notice buried in a developer newsletter nobody in your business subscribes to. Somebody has to read those notices and act on them. A missed deprecation stops the integration outright, and it always happens at the least convenient moment.
Schema drift on your own side
Your CRM administrator adds a required field. Your finance team renames a category. Neither knows an integration depends on the old shape. Most integration failures in year two originate internally rather than at the vendor, which is why lightweight change control matters more here than elegant code.
Monitoring and support cover
Somebody must be accountable for a failure at four o’clock on a Friday. Whether that is a named internal owner or a supplier retainer, write it down with a response time attached. Unowned integrations decay predictably, and the decay is usually discovered by a customer rather than by a dashboard.
Small enhancements and tuning
Nobody gets the mapping perfect first time. Expect a steady trickle of small changes — an extra field, a different filter, a new status value — and set aside two to four days a year for them. Left to accumulate, that trickle becomes a project with its own API integration cost and its own approval cycle.
In-house versus outsourced API integration cost
This comparison is usually made on day rate alone, which is the wrong axis entirely. What matters is who carries the risk of the unknown, and who will still be available in eighteen months when the vendor changes an endpoint without much warning.
Building with your own developers
Cheaper per hour, and better over the long run if you already employ people who know both systems. The hidden costs are opportunity cost — the roadmap work that does not happen — and the bus factor. One developer who understands the integration, with no documentation behind them, is a liability rather than a saving.
Using a specialist partner
An experienced software development partner has met the vendor’s API before, and that experience is what the premium actually buys. Insist on named people, a written estimate that separates discovery from build, and source code in your own repository from the first commit rather than at the end of the engagement.
The blended model that usually wins
Let the partner lead discovery and the first integration while your team pairs on it and takes over maintenance afterwards. It costs slightly more than pure outsourcing in year one and materially less by year three, and it keeps the knowledge inside the business where the next API integration cost decision will be made.
Offshore rates and the total figure
Day rates of £150–£300 are genuinely available and sometimes work very well. Account honestly for the coordination overhead: time-zone overlap, the precision your specification now requires, and review time from somebody senior on your side. A well-run offshore build lands around thirty per cent cheaper overall, not seventy.
How to estimate your API integration cost in one afternoon
You do not need a supplier to reach a defensible number. The method below gets most businesses within twenty per cent of the final API integration cost, which is enough to decide whether to proceed and enough to interrogate a quote intelligently.
Count endpoints, not systems
“Integrate the CRM with the ERP” is not a scope. “Create a customer, update a customer, create an order, read stock levels, write invoice status” is five endpoints. The endpoint is the unit of work, because each one carries its own mapping, error handling, testing and documentation.
Score each endpoint for complexity
Give every endpoint a one for simple (few fields, one direction), a two for moderate (transformation or lookups required) or a three for complex (bidirectional, conflict rules, or high volume). Sum the scores, then multiply by roughly £900–£1,400 a point at UK agency rates to reach a working API integration cost.
Add the fixed project costs
Discovery, monitoring setup, deployment and documentation do not scale with endpoint count. Add £2,500–£5,000 for the project regardless of size, plus any sandbox fee or vendor licence-tier upgrade the integration requires. These fixed items are what make very small integrations look disproportionately expensive per endpoint.
Apply contingency by unknown
Twenty per cent on a documented, modern API you have actually read; thirty to forty per cent on a legacy or undocumented system. If you cannot see the documentation before quoting, that contingency is not padding — it is the price of information you do not yet have, and it should be stated as such.
Worked example: e-commerce to accounting
Five endpoints scoring one, two, two, three and one gives nine points. At £1,100 a point that is £9,900, plus £3,500 of fixed costs, plus twenty per cent contingency: about £16,000, with maintenance of £2,400–£4,000 a year. That is a credible API integration cost for a mid-market bidirectional sync, and a £4,000 quote against the same scope should worry you rather than please you.
Reducing the bill without cutting corners
There is a considerable difference between spending less and buying less. The moves below reduce an API integration cost genuinely, whereas cutting monitoring, error handling or documentation merely defers the spend and adds interest to it.
Ship one direction first
Deliver the direction carrying the business value, run it for a month, then add the return path if it is still wanted. A surprising number of two-way requirements turn out to be one-way plus a report — the same interrogation the build vs buy question deserves before any budget is committed.
Reduce the field list
Every field is mapping, testing and a future failure point. Most integrations move perhaps sixty per cent of the fields somebody originally requested. Ask which decision each field supports; where nobody can answer, leave it out of the first release and revisit it once the pipe is running in production.
Evaluate the native connector first
Where both platforms publish a native connector, assess it honestly before commissioning anything. It will be less flexible and it may not cover your edge cases, but a zero build cost with vendor-maintained upgrades beats elegance for a genuinely commodity link between two mainstream products.
Batch instead of streaming
Real-time is the most expensive requirement in integration work. If the business can live with a fifteen-minute cycle, the architecture simplifies, the rate-limit problem largely disappears, and the API integration cost typically drops by around a third with no loss anybody notices.
Consolidate onto shared patterns
Your second and third integrations should cost less than the first, but they only will if the first established reusable patterns for authentication, logging and retries. That is the argument for a little technology consulting and DevOps discipline up front instead of three unrelated scripts maintained by three different people.
Frequently asked questions about API integration cost
How much does a single API integration cost?
For one straightforward, one-directional connection between two modern SaaS platforms, £2,000–£6,000 is the normal UK range, delivered in two to four weeks. Bidirectional sync with conflict rules pushes the API integration cost to £8,000–£20,000. Anything spanning three or more systems should be scoped and funded as a platform, not as a task.
Why do quotes for the same integration vary so widely?
Because they are not pricing the same scope. A quote including discovery, error handling, monitoring and documentation always looks expensive beside one covering the happy path alone. Ask both suppliers to price those five components explicitly and the apparent gap in API integration cost usually collapses to something small.
Is an integration platform cheaper than custom code?
At low volumes and for simple connections, almost always. As volume, complexity or the number of integrations rises, per-task pricing and platform licences overtake the cost of code you own outright. Model both routes across three years rather than comparing month one, and include the migration cost of leaving.
What should I budget for ongoing maintenance?
Fifteen to twenty-five per cent of the build cost a year, or a retainer of £800–£2,500 a month across a small estate. That covers deprecations, schema changes, monitoring and small enhancements. Budgeting nothing for maintenance is the most common decision here and reliably the most expensive one.
Can we reduce the API integration cost by building it in-house?
Sometimes, if you employ developers who know both systems and can genuinely spare them. Compare against opportunity cost and be honest about who owns documentation and maintenance afterwards. The blended approach — a partner for the first build, in-house for the rest — usually produces the lowest three-year figure.
How long does an API integration take to deliver?
Two to four weeks for a simple one-way push, six to ten weeks for bidirectional sync, and three to six months for multi-system orchestration delivered in phases. Vendor sandbox access and security review cause more delay than development does, so start both conversations in week one.
Turning the estimate into a decision
An API integration cost only becomes useful when it is attached to a decision: proceed, phase, or use the vendor connector and revisit next year. Do the endpoint count, apply the contingency that matches how much you actually know, and put the maintenance percentage in the same paper as the build figure.
Ask every supplier the same five questions
Which endpoints are in scope; how are errors handled and replayed; what monitoring is included; what happens when the vendor changes the API; and what does year two cost. Suppliers who answer those clearly are pricing the whole job, and their number is comparable with anybody else’s.
Fund maintenance in the original business case
The integration that fails in year two is almost never the one that was built badly. It is the one nobody was paid to look after. Naming an owner and a budget line at approval time costs nothing and prevents the most expensive outcome available to you.
Keep the estimate as a living document
Revisit the endpoint list and the assumptions when discovery finishes, and again at go-live. An API integration cost that was honest in January and never updated is how projects drift quietly past their approval without anybody making a decision to let them.
