reCAPTCHA v3 solver API

Solve request#

Solving reCAPTCHA v3 is a single synchronous call. You post one task object and the response carries the finished solution — no task id to store, nothing to poll for.

The connection is held open for the whole solve, which for this type runs a median of 285ms. Set a client timeout that covers it — the default socket timeout in most HTTP libraries is shorter than the work. If holding a connection that long does not suit your worker fleet, the same task goes to /v1/createTask and you collect it from /v1/checkTask instead.

When to use it#

• Load-testing a checkout that scores every request
• Monitoring conversion funnels that silently gate on a score threshold
• Verifying your own score thresholds behave as configured

Task object structure#

These fields go inside task. The envelope around it — clientKey and the optional softId — is the same for every type and is described in the API reference.

Example payload#

The whole body, assembled from the table above — copy it formatted to read, or on one line to paste straight into a terminal or a REST client. The values are placeholders: swap the key for your own and the URL and site key for the page you are solving.

{
  "clientKey": "void_a1b2c3d4e5f60718293a4b5c6d7e8f90",
  "task": {
    "type": "ReCaptchaV3",
    "url": "https://example.com/login",
    "site_key": "6LdyC2cUAAAAACGuDKpXeDorzUDWXmdqeg-xy696",
    "action": "login",
    "proxy": "proxy.example.com:9090:user:pass"
  }
}

Example request#

The smallest thing that gets a token back, in your language. Python and Node.js use the official client (pip install voidcaptcha, npm install voidcaptcha); every other language makes the HTTP call its standard library already has. The complete version, with the key read from the environment and the error envelope checked, is in Sample code at the bottom of this page.

Language choice is remembered across the whole site, so switching here carries to every other sample in the docs.

# pip install voidcaptcha
from voidcaptcha import VoidCaptcha

client = VoidCaptcha("void_a1b2c3d4e5f60718293a4b5c6d7e8f90")

result = client.recaptcha_v3(
    url="https://example.com/login",
    site_key="6LdyC2cUAAAAACGuDKpXeDorzUDWXmdqeg-xy696",
    action="login",
    proxy="proxy.example.com:9090:user:pass",
)

print(result.value)

Example response#

A successful solve returns errorId: 0 and a solution object. The result is in solution.token; alongside it is the userAgent the solve ran under.

solution carries those and nothing else — the gateway copies the result field, anything extra this type returns, and userAgent across, and drops the rest of whatever the solver sent. Timing arrives at the top level as elapsedMs, present only when the duration is known, beside cost and your remaining balance. The figures below are illustrative: nothing here is a published median.

{
  "errorId": 0,
  "status": "ready",
  "solution": {
    "token": "03AGdBq25V6…",
    "userAgent": "Mozilla/5.0 …"
  },
  "cost": "$0.0008",
  "balance": "$411.24",
  "elapsedMs": 4820
}

Send that userAgent on your downstream request as well — for the cookie-based types the result is bound to it and will be rejected under a different one.

Errors for this type#

The codes reCAPTCHA v3 can return, on top of the shared authentication and billing set. Branch on errorCode, not on the HTTP status.

Every code, with the fix for each, is on the error reference. A failed solve is never charged.

Sample code#

A complete file: key from the environment, task built, error envelope checked, result used. It calls the endpoint directly rather than through an SDK, so it works as a starting point in any language.

import os
import sys

import requests

API = "https://api.voidcaptcha.com/v1/getResult"

payload = {
    "clientKey": os.environ["VOIDCAPTCHA_KEY"],
    "task": {
        "type": "ReCaptchaV3",
        "url": "https://example.com/login",
        "site_key": "6LdyC2cUAAAAACGuDKpXeDorzUDWXmdqeg-xy696",
        "action": "login",
        "proxy": "proxy.example.com:9090:user:pass"
    },
}

# The endpoint is synchronous: it holds the connection open for the whole
# solve, so the read timeout has to cover it, not just the handshake.
res = requests.post(API, json=payload, timeout=180).json()

if res["errorId"] != 0:
    sys.exit(f"{res['errorCode']}: {res['errorDescription']}")

solution = res["solution"]

# Carry the User-Agent downstream — the result was earned under it.
print(solution["token"])
print(solution["userAgent"])