Skip to content
NexCaptcha

Migration guide

Switching from CapMonster Cloud

CapMonster is cheaper per 1,000 on reCAPTCHA v2 and images. The difference is what happens on a miss, and which types are on the published task list.

API compatibilityDrop-in — change the base URL

What they do well

  • Cheaper than us on the types most people send most, and we are not going to pretend otherwise: reCAPTCHA v2 at $0.60 per 1,000 against our $1.50, reCAPTCHA v2 Enterprise at $1.00 against our $1.50, Turnstile at $1.30 against our $2.18, and grid/click image recognition from about $0.04 per 1,000 images against our $0.75. If unit price is the only variable you are optimising, they win that comparison outright.
  • Genuinely fast, with pure ML and no human queue in the path. Their own documented cycle is createTask, wait 300 ms to about 6 seconds, then getTaskResult, and an independent Trustpilot reviewer running it beside Anti-Captcha measured 1 to 3 seconds against 10 to 20.
  • The best migration tooling in the category, and they built it deliberately: documented compatibility endpoints and an interceptor covering Anti-Captcha v1 and v2, RuCaptcha, RipCaptcha, 2Captcha, BypassCaptcha and DeathByCaptcha, plus a hosts-file remap so GUI automation software can keep calling the old hostname.
  • Built by ZennoLab, an established automation vendor, with first-party SDKs in Python, .NET, JavaScript, Go and Java, and consumer terms better than most here: no minimum deposit, a $0.10 test balance on request, a three-day refund window, and billing only on solved captchas.
  • Trustpilot 4.6 out of 5 across 106 reviews, 96% five-star, with the praise concentrated on exactly the things they optimise for — speed, price and API ergonomics.
  • Built-in proxy infrastructure on most task types, so a lot of integrations never manage a proxy pool at all.

Where they fall short

  • No human fallback anywhere in the stack. When the model has not been trained on a target's current challenge variant the task simply fails — it does not degrade into a slower but correct answer. That shows up as sudden total failure on one specific site rather than gradual drift, which is the structural cost of the speed advantage.
  • DataDome is the documented sore spot. An April 2026 Trustpilot reviewer reports it not solving DataDome at all and describes support answering with copy-pasted messages.
  • Support responsiveness is the single most consistent complaint theme, and it spans years: reviewers in December 2025 and March 2025 both describe sending the email form and then waiting weeks without a useful reply.
  • Task naming is out of step with the rest of the ecosystem. reCAPTCHA v2 is NoCaptchaTask, a name inherited from Antigate v1 that nobody else still uses, so CapMonster code is not copy-paste portable to 2Captcha v2 or CapSolver without a rename pass.
  • Coverage is mid-pack. Strong on reCAPTCHA, Turnstile and image work; thin on the long tail. Tencent, KeyCaptcha, Capy, MTCaptcha and Lemin are not on their documented task list, and neither is the full Cloudflare interstitial.
  • capmonster.cloud refused TCP connections on port 443 from two independent clients when we checked prices on 2026-09-22, which left their own pricing page unreadable from here. We record that as a dated observation, not as a claim about their uptime — but if you cannot load a vendor's price list from your own network, that is a procurement problem.
  • The three-day refund window is short if you are evaluating against a real production workload rather than a smoke test.

[ 02 ]Side by side

Side by side

Side by side
CapMonster CloudNexCaptcha
Published pricingreCAPTCHA v2 $0.60, Turnstile $1.30, images from ~$0.04 per 1,000. Cheaper than us.reCAPTCHA v2 $1.50, Enterprise $1.50, Turnstile $2.18, image-to-text $0.75. 33 types, all flat.
Published latencyDocumented cycle of 300 ms to about 6 s. No percentiles, but that number is theirs.p50 and p95 published per type. Overall p50 4.8s, p95 11.2s.
Published success rateNot published, and there is no human fallback when the model misses.Published per type, 99.4% overall, counted on tokens the target accepted.
What counts as a solveA solved captcha per their FAQ. Model misses fail outright rather than degrading.The target site accepting the token. We bill on that, not on returning one.
Failed-solve billingOnly successful solves are billed. Deposit refunds within three days of purchase.Rejected, failed and timed-out solves are refunded automatically within the hour.
API compatibilitycreateTask/getTaskResult, plus compatibility endpoints for six other vendors.Same envelope and callbackUrl. One rename: NoCaptchaTask becomes RecaptchaV2Task.
Support responseEmail form only. Slow replies are the most repeated complaint, across several years.24/7 engineers. Median first response 7 minutes on Telegram.

[ 03 ]The code change

The code change

CapMonster reimplements the Anti-Captcha convention, which is also what we implement: JSON POST to /createTask, /getTaskResult and /getBalance, clientKey in the body, and the {errorId, taskId} then {errorId, status, solution} envelopes. callbackUrl keeps its name and its payload shape, so a webhook handler written for them works here untouched. The one place the two vocabularies diverge is reCAPTCHA v2, which they still call NoCaptchaTask — that is the single rename in the diff below and in step 2.

Before
const task = { type: "RecaptchaV3TaskProxyless", websiteURL, websiteKey, minScore: 0.7 };
const res = await fetch("https://api.capmonster.cloud/createTask", {
method: "POST",
headers: { "content-type": "application/json" },
body: JSON.stringify({ clientKey: API_KEY, task }),
});
After
const task = { type: "RecaptchaV3TaskProxyless", websiteURL, websiteKey, minScore: 0.7 };
const res = await fetch("https://api.nexcaptcha.net/createTask", {
method: "POST",
headers: { "content-type": "application/json" },
body: JSON.stringify({ clientKey: API_KEY, task }),
});

How to migrate

  1. 01Point your existing client at https://api.nexcaptcha.net/createTask and swap the clientKey value — the paths and the request envelope are identical.
  2. 02Rename the reCAPTCHA v2 task type: NoCaptchaTask becomes RecaptchaV2Task and NoCaptchaTaskProxyless becomes RecaptchaV2TaskProxyless. For most CapMonster clients this is the only rename in the whole migration; RecaptchaV3TaskProxyless, RecaptchaV2EnterpriseTask, TurnstileTask, GeeTestTask, ImageToTextTask, DataDomeTask, ImpervaTask and AmazonTask all carry over as they are.
  3. 03Map ComplexImageTask by hand. It folds grid, click and bounding-box work into one type selected by a class field, and those land on separate types here — click-captcha, slider-captcha and image-to-text, at $1.80, $1.80 and $0.75 per 1,000.
  4. 04Loosen your polling deadlines before you cut over. CapMonster clients are usually written around a 300 ms to 6 second cycle; our overall p50 is 4.8s and p95 11.2s, so a naive port with a 6 second deadline will look like a regression on day one. Per-type percentiles are published — use the ones for the types you actually send.
  5. 05If you drive CapMonster through CapMonsterInterceptor or the hosts-file remap rather than calling the API directly, undo the remap and point the original software at api.nexcaptcha.net.

[ 04 ]Questions about this one

Questions about this one

We are, on the types you send most — $1.50 against $0.60 on reCAPTCHA v2 is two and a half times. The case for moving is not the rate, it is what a failure costs you. CapMonster has no human fallback, so a challenge variant their model has not seen fails outright on that one site until they retrain, and their most-cited weak spot is DataDome. If a blocked target costs you more per hour than the price delta costs you per month, the cheaper token is the more expensive decision. If it does not, stay where you are.

One of them. reCAPTCHA v2 is NoCaptchaTask on CapMonster, inherited from Antigate v1, and it is RecaptchaV2Task here. Everything else you are likely to be sending — RecaptchaV3TaskProxyless, RecaptchaV2EnterpriseTask, TurnstileTask, GeeTestTask, ImageToTextTask, DataDomeTask, ImpervaTask, AmazonTask — keeps its name. ComplexImageTask is the other one to look at, because it covers three of our types depending on its class field.

On the fast types, probably slightly. CapMonster documents a 300 ms to 6 second cycle and our overall p50 is 4.8s with p95 11.2s. Per-type is the honest comparison: Turnstile is p50 3.1s here, image-to-text p50 1.9s, and reCAPTCHA v2 p50 4.2s. Raise any 6 second polling deadline before you cut over, then compare accepted rate rather than time to first response.

Yes. Balances cannot move between vendors in this market, so the sensible pattern is a percentage split in your client: send the types you have had trouble with here, keep the cheap high-volume reCAPTCHA v2 traffic there while you spend the balance down, and decide per type instead of all at once. Plenty of customers end up running both permanently, and that is a reasonable outcome.

Solve your first thousand on us.

Create a key, paste it into the sample above, and watch a real token come back. No card, no sales call.

Get your API keyTalk to an engineer