Crypto AML Screening Cost: Build vs Buy for KYT
Crypto AML screening cost for EU CASPs covering vendor list vs negotiated pricing, a real production custom-stack's launch economics, the hidden engineering costs vendor quotes leave out and the build-vs-buy decision factors that actually matter.
12 min read
56 views
Key takeaways: crypto AML screening cost 5
The cost and build-vs-buy decisions that matter most when budgeting a KYT/AML screening integration for an EU CASP, from vendor list-vs-negotiated pricing through a real custom stack's launch economics to when buying still wins.
- List price is not real price Enterprise KYT contracts negotiate well off list; Chainalysis list is €120K-250K/year, negotiated contracts reported by Vendr and Costbench average around $174.7K, ranging $25.7K-$297.3K.
- A minimal custom stack can launch for $35-70/month CoinGecko's $35/month tier is the only unavoidable payment, but the figure excludes engineering time and is gated on unobtained commercial-use permissions.
- Pagination and multi-hop tracing are the real engineering cost Whale-wallet checks scale toward $0.05/check on pagination alone, multi-hop tracing at high branching factor runs into millions of compute units, and a properly-keyed cache cuts external costs 30-60%.
- Detection quality is the honest center of build vs buy FP/FN rates for a custom stack are explicitly unmeasured; buying a vendor's track record is the strongest liability-side argument, and building earns its keep as a complement rather than a full replacement for most launch-stage CASPs.
- The right default depends on stage and scrutiny Pre-launch CASPs default to the cheapest viable option, growth-stage teams should compare against negotiated (not list) vendor pricing, and heavy regulatory scrutiny favors a source-traceable audit trail.
Every EU CASP evaluating a KYT vendor eventually asks the same question: what is the real crypto AML screening cost, not the number on a pricing page. Vendor pages answer with a monthly figure or a "contact sales" wall, and neither tells you what a working screening stack actually costs to run once integration engineering, infrastructure and data-quality risk are added in. This guide breaks that total down using figures from an anonymized production build we delivered for a real EU CASP evaluating and prototyping a replacement for its incumbent screening vendor, part of our Web3 FinTech development practice.
In short: a list price or per-check fee says little about the real cost of crypto AML screening. The real total is vendor fees plus integration engineering plus infrastructure plus the risk cost of data quality, and a minimal custom stack can run for tens of dollars a month at launch volume while buying unmeasured detection quality and an ongoing engineering burden in return. For most CASPs, build vs buy is mostly a liability and time-to-market decision, not a pure cost comparison, and what follows is hedged accordingly.
What crypto AML screening actually costs
The vendor's fee is the line everyone budgets for: the check itself, priced per transaction, per wallet or as a flat subscription. Underneath it sit costs that rarely make it onto a budget line: the engineering time to wire that API, or a custom stack, into a deposit or withdrawal flow, the infrastructure that keeps a screening call off a synchronous request path, the quieter cost of data quality (a screening engine that misses a pattern or over-flags a legitimate customer carries a cost, in regulatory exposure on one side and customer friction on the other, that a subscription fee never itemizes), plus the compliance overhead of reviewing what a verdict actually means for a specific transfer. The table below breaks down what drives each component.
| Cost component | What drives it |
|---|---|
| Vendor fees | Per-check or per-wallet pricing, tier minimums, bundled call caps and enterprise contract terms |
| Integration engineering | Wiring the vendor API, or a custom stack, into deposit and withdrawal flows, orchestration and a verdict-response workflow |
| Infrastructure | Compute-unit metered data providers, caching layers and the servers a custom stack needs to run continuously |
| Data quality risk | Coverage gaps, stale feeds and unmeasured detection quality, none of which are a line item but all of which carry downstream risk |
| Compliance overhead | Reviewing flags, maintaining an audit trail and demonstrating control adequacy to a supervisor |
Vendor pricing models: list price vs real price
KYT vendors do not price on one model. At the lightweight, self-serve end, AMLBot's Lite plan is quoted from $9 for 20 checks, close to $0.45 per check, with subscriptions starting at $229 or more a month. MistTrack runs a tiered dashboard subscription, Basic at $229 a month (up to 50 addresses, no API), Standard at $689 a month (API at one call per second), Compliance at $2,069 a month (five calls per second), alongside a separate prepaid volume ladder from $0.20 a call down to $0.05 a call at its largest package.
Enterprise vendors quote a claimed or list price, and market-intelligence data suggests negotiated contracts run both below and above that list. Chainalysis lists an estimated €120,000 to €250,000 a year; negotiated contracts reported through Vendr and Costbench range $25,700 to $297,300 a year, averaging around $174,700, with large deals scaling past $500,000 to $1,000,000 or more a year. Elliptic's claimed range is €80,000 to €180,000 a year against reported negotiated contracts of $150,000 or more; TRM Labs' entry point is quoted at €60,000 to €150,000 a year, with no separately-reported negotiated figure available. These are figures reported through public procurement-intelligence data, not a contract any specific client signed, and reaching this class of coverage independently is estimated at $50,000 to $500,000 or more a year overall.
The model itself varies as much as the number: per check or credit, flat tiered subscription with a call cap, an abstracted resource-consumption unit, or per successful query with a volume discount. Comparing "per-check price" across vendors often compares figures that are not measuring the same unit.
The custom-stack launch economics
Set against those vendor numbers, the custom stack built for the anonymized CASP project behind this guide has a small launch-stage bill. At launch volume, on Ethereum and Base, running hundreds of checks a month, the project's own cost summary puts outside data services at about $35 to $70 a month. CoinGecko's Basic fiat-price tier at $35 a month is the only genuinely unavoidable payment; the rest is free or optional: Alchemy chain data free up to 30M compute units a month (roughly 60,000 checks); public sanctions lists and GraphSense exchange labels free; ScamSniffer free with a seven-day delay or $999 a month real-time; Chainalysis' free Sanctions API capped at 5,000 requests/5 minutes, internal use only; MistTrack per-call enrichment roughly $20 to $30 a month, the optional half of the range; TRM's free Sanctions API as a backup; BigQuery's public Ethereum dataset free for later batch work, not live screening.
Add person screening, sanctions and PEP matching alongside the on-chain checks, and the figure moves to about $50 to $120 a month, mainly an OpenSanctions subscription around EUR 10 to 50 a month. Against AMLBot-class subscriptions starting at $229 or more a month, and enterprise data at $50,000 to $500,000 or more a year, the launch-stage custom stack looks like an easy call on price alone.
It is not the full picture. The $35-70 figure assumes commercial-use permissions, letters covering Alchemy white-labeling, CoinGecko price storage, MistTrack embedding and OpenSanctions reselling, are actually obtained; those are gates, not formalities. It excludes engineering time to build and maintain the stack, and it says nothing about detection quality: how reliably the custom stack catches illicit activity is not measured, and a validated reference set large enough to measure it against does not yet exist. That gap, not the monthly bill, is the honest center of the build-vs-buy decision. A fully-loaded production version of the same stack, engineering headcount, commercial data feeds, hosting, legal support and liability insurance included, runs closer to $17,000 to $28,500 a month, roughly $204,000 to $342,000 a year, a figure the source material keeps deliberately separate from the launch number above.
The hidden engineering costs nobody quotes
None of the numbers above account for what it costs to make a check finish. An ordinary address check runs roughly 500 compute units, well under half a cent, but a heavy exchange-style wallet check scales toward roughly $0.05 a check purely from pagination volume: a whale-scale wallet with on the order of a million transfers across roughly a thousand pages takes tens of minutes to retrieve in full even at close to two seconds a page. That per-check cost also marks the free-tier boundary: at roughly 500 compute units per ordinary check, the 30 million compute unit monthly free tier covers roughly 60,000 checks, which is where the launch-stack economics above stop holding and vendor or paid-tier pricing takes over. Live multi-hop tracing is worse: a breadth-first walk at a branching factor of 10 costs roughly 26,640 compute units for a three-hop check, and at a branching factor of 100, plausible at an exchange router or a dusting fan-out, the cost explodes to roughly 2.42 million. A properly-keyed result cache is reported to cut external data-provider costs by roughly 30 to 60 percent, the single biggest lever before touching the vendor relationship at all. This is the same architecture problem covered in more engineering depth in our crypto transaction monitoring integration guide, a companion piece to this one.
Build vs buy: the real decision factors
Coverage and data freshness
Cost alone does not resolve the build-vs-buy question, the two paths differ on coverage, liability and speed in ways a monthly bill does not capture. On coverage, buying wins by default: a proprietary label database with hundreds of millions of tags dwarfs an open-source library's roughly 500,000, and enterprise vendors cover 15 to 24 or more networks against a launch-stage build's two EVM chains. Building narrows that gap only if traffic is concentrated: the anonymized client's own split ran roughly 95 percent Ethereum, 4 percent Base and close to zero elsewhere, so a narrow EVM-only build already covered around 99 percent of real volume, a result that does not generalize to genuinely diversified multi-chain traffic.
Both paths carry a freshness and transparency cost: ScamSniffer's free feed has a seven-day delay against $999 a month real-time, and open PEP or adverse-media data is materially weaker than commercial databases. In build's favor, a custom stack can be evidence-based, every verdict traces to a named source, date and snapshot, versus a vendor's opaque score, a difference the source project ties to the AMLR's Article 77 and Article 20 auditability requirements.
Liability and regulatory exposure
The honest answer on liability cuts against build. A naive one-hop custom build without deposit-clustering is, in the source project's own cross-engine assessment, "a transparent box that misses crime instead of a black box that finds it", and live multi-hop tracing over a synchronous path is independently flagged mathematically unsafe at realistic branching factors. Detection-quality rates for the custom stack are explicitly unmeasured either way, and a validation set large enough to change that does not exist yet, the single largest liability argument for buying a track record instead of building one. Narrowly in build's favor: benchmarking a live incumbent account risks violating a vendor's terms of service against "competitive analysis"; owning the stack removes that exposure without answering the detection-quality question.
Time to market and engineering cost
On time to market, a full open-source label and graph self-build runs 4 to 6 months with 2 to 3 engineers and an analyst. A white-label partnership integration lands in 1 to 2 months at tens of thousands of dollars a year, and a bounded EVM-native hosted-API prototype, the launch-stage scope costed above, ships in roughly 2 to 4 weeks with multi-hop tracing deferred to later. On engineering cost the paths cross over by volume: at high volume, closing the gap to enterprise-vendor parity costs the same annual range as above, and self-hosting a comparable infrastructure class runs roughly $2,000 to $3,500 a month bare metal before licensing; at low volume vendor minimums make a $35-70 marginal launch cost favorable, buying long-tail data incrementally instead of renting it upfront. Switching either direction carries its own cost too, a parallel-run period, re-screening the existing customer back book and porting historic verdicts and audit trails.
| Factor | Favors buy | Favors build |
|---|---|---|
| Coverage and data quality | Broader network coverage and larger label databases out of the box | Traffic concentrated on 1-2 chains, where a narrow build already covers most real volume |
| Liability and regulator optics | A vendor's track record, versus an unmeasured custom detection rate | Avoids ToS exposure from benchmarking a live incumbent account |
| Time to market | White-label integration in 1-2 months | A bounded prototype ships in weeks, a full self-build takes 4-6 months |
| Engineering cost | Long-tail data parity independently costs $50K-500K+ a year at high volume | A marginal launch cost near $35-70 a month is favorable specifically at low volume |
| Flexibility and vendor lock-in | No engineering burden to build and maintain an evidentiary layer yourself, though switching later carries its own re-screening and audit-porting cost | Every verdict traces to a named source, date and snapshot instead of a vendor's opaque score |
For most launch-stage CASPs this points toward buying for coverage and liability, and treating a custom stack as a complement rather than a full replacement: pre-screening, caching and routing in front of a vendor, not instead of one, is the pattern we recommend once the detection-quality gap is priced in honestly.
Cost decision guide
None of the above resolves into one universal answer. The following is a set of defaults to reason from, not a rule to apply without adjustment.
| Situation | Default | Note |
|---|---|---|
| Pre-launch CASP scoping its first screening integration | Start with the $35-70/month launch-stage custom stack or a lightweight vendor tier, not an enterprise contract | Enterprise data at $50K-500K+ a year is a high-volume decision, not a launch cost |
| Growth stage with volume-driven vendor bills | Reassess against negotiated, not list, enterprise pricing before assuming the vendor is the expensive option | See the market-reported negotiated figures above, list price alone overstates what CASPs actually pay |
| Multi-chain expansion beyond 1-2 EVM chains | Weight coverage more heavily, a narrow custom build's cost advantage shrinks as chain count grows | The roughly 99% coverage a narrow EVM build reached does not generalize past a concentrated traffic split |
| Heavy regulatory scrutiny or a supervisor asking for evidentiary depth | Favor the option with a source-traceable audit trail per verdict | AMLR Article 77 requires unredacted, admissible records, a black-box vendor score is harder to defend on its own |
| Hybrid path | Default: buy for coverage and liability, build a pre-screening and caching layer in front of it | A properly-keyed cache alone is reported to cut external data costs by 30-60% |
How Pharos Production helps with screening economics
The numbers above draw on an anonymized real-project evaluation: an EU CASP scoping and prototyping a replacement path for its incumbent crypto-AML screening vendor, weighing vendor pricing against a minimal custom stack's launch economics and the liability trade-offs above. We do not publish a detection-quality number for either path: no published false-positive or false-negative rates exist for the approaches described and a validated reference set does not exist for this stack. What we can help with is the part that is solvable: pricing out the real total cost of a screening integration for your CASP's actual chain mix and volume, scoping the engineering effort honestly against a vendor's list and negotiated pricing and building whichever mix of buy and build makes sense, as part of a custom software development engagement.
Sources: an anonymized production EU CASP vendor-evaluation and build-vs-buy project, cross-checked against public KYT vendor pricing pages and market-reported negotiated contract figures (Vendr, Costbench). No detection-quality, false-positive or false-negative figures are published here for any approach; that measurement does not exist in the source material and remains unsolved.
FAQ
Quick answers to common questions about custom software development, pricing, process and technology.
Type to filter questions and answers. Use Topic to narrow the list.
Showing all 5
No matches
Try a different keyword, change the topic or clear filters
-
Costs range widely by tier. Lightweight vendors start around $0.45 per check or $229 a month or more; enterprise-grade coverage lists at €60,000 to €250,000 a year depending on the provider, though negotiated contracts run both lower and higher than list price (Chainalysis negotiated contracts reported by Vendr and Costbench range from $25,700 to $297,300 a year, averaging around $174,700).
A minimal custom-built stack can run about $35 to $70 a month at launch volume, before engineering time is added in.
-
Per-check pricing is really only common at the lightweight end: AMLBot's Lite tier works out to roughly $0.45 a check. Most vendors price by subscription tier instead, MistTrack's dashboard runs $229 to $2,069 a month depending on API access and call rate, with a separate pay-as-you-go ladder that drops from $0.20 to $0.05 a call at volume.
Enterprise players skip per-check pricing entirely and sell annual contracts instead.
-
Yes, at least for a narrow scope. A production EU CASP project prototyped a minimal EVM-native screening stack (Ethereum and Base) using free and low-cost data sources, Alchemy for chain data, CoinGecko for fiat prices, public sanctions lists and open-source label sets, shipping in roughly 2 to 4 weeks.
What that scope does not include is multi-hop tracing, broad multi-chain coverage or measured detection quality, all separate, larger engineering efforts.
-
On the pure data-cost line, yes at launch volume: the source project's launch-stage custom stack ran about $35 to $70 a month in outside data costs, against vendor minimums starting around $229 a month for lightweight tiers and mid-market entry points around €60,000-150,000 a year. That comparison leaves out engineering time to build and maintain the stack and, most importantly, leaves detection quality unmeasured, so "cheaper" describes the data bill only, not the full cost or the risk trade-off.
-
Beyond the vendor invoice, the biggest cost drivers are integration engineering, wiring screening into a deposit or withdrawal flow, infrastructure for compute-metered data providers and caching, plus compliance overhead for reviewing flags and maintaining an audit trail. At the data layer, whale-wallet pagination and live multi-hop tracing scale compute-unit costs sharply, a three-hop trace at a high branching factor can run into the millions of compute units per check, which is why a properly-keyed cache, reported to cut external data costs by 30 to 60 percent, matters more to the real bill than the vendor's list price.
I work with startup founders who need a dedicated software development team but don’t want to gamble on hiring, random outsourcing, or opaque delivery.
Most founders face the same problem sooner or later.
Early technical and team decisions lock the product into tech debt, slow delivery, missed milestones and constant re-hiring. By the time this becomes visible, fixing it is already expensive.As a CTO and software architect, I help founders design, build and run dedicated development teams that work as a true extension of the startup. Not as a black-box vendor.
My focus is on complex products where mistakes are costly:
- Web3 and blockchain platforms
- FinTech and regulated products
- High-load startup systems
- MVP → scale transitions
We don’t do body-shopping.
We don’t sell generic outsourcing.Instead, we help founders:
- build the right team structure from day one
- keep technical ownership and transparency
- scale delivery without losing control
- avoid vendor lock-in and hidden risks
Teams are aligned with the product roadmap, business goals and long-term architecture. Not just short-term velocity.
