GeeTest 求解器， 返回即得令牌。

无需浏览器的 GeeTest 求解 API。Takion 一次调用即可返回有效的 GeeTest 解——覆盖 v3 和 v4，滑块、图标和 AI 验证码——让你的表单顺利提交、请求顺利通过。

最后更新 · 覆盖能力已针对一个实时的 GeeTest 目标测试。

被 GeeTest 拦住了？

GeeTest 会给站点一道验证码——滑块拼图、图标网格或 AI 挑战——并期待收到一套匹配的令牌。Takion 在服务端解决它，并返回相应的值（v3：geetest_validate；v4：lot_number、pass_token、captcha_output），无需浏览器，也无需拖动拼图。把它们随表单一并提交，检测即告通过。

GeeTest 的工作原理

GeeTest 是一种行为验证码，广泛用于登录、注册和结账流程。它提供多种挑战类型——滑块拼图、图标点选网格和基于 AI 的验证码——分为两代：v3 返回 geetest_challenge / geetest_validate / geetest_seccode，v4 返回 captcha_id / lot_number / pass_token / captcha_output。大多数站点仍在使用 v3，部分已迁移到 v4。

Takion 在服务端解决挑战，并针对任一版本返回站点用于校验的整套令牌。一次调用，无需浏览器，也无需拖动拼图。你按站点所需的方式，把返回的各项值随表单或请求一并提交即可。

  • Takion 覆盖的挑战
  • 滑块验证码
  • 图标／点选验证码
  • AI／空间验证码
  • GeeTest v3 和 v4

一次调用返回什么

向 GeeTest 端点发送一次调用，即可针对 v3 或 v4 返回站点用于校验的确切令牌集。

captcha_output

v4 的解，含 lot_number、pass_token 和 captcha_id

geetest_validate

v3 的解，含 geetest_challenge 和 geetest_seccode

curl -G https://geetest.takionapi.tech/v4 \
  --data-urlencode "api_key=$TAKION_KEY" \
  --data-urlencode "captcha_id=fcd636b4514bf7ac4143922550b3008b" \
  --data-urlencode "kind=slide"
# → { "captcha_id": ..., "lot_number": ..., "pass_token": ..., "captcha_output": ... }

阅读 GeeTest 文档同一个密钥，通吃每一堵墙。同一种 JSON 结构，适配每一个厂商。

How the GeeTest solve works, end to end

No browser, no rendered puzzle. Four moves from challenge to token.

  1. 1

    You read the challenge params

    Pull them from the page HTML or the captcha's init endpoint. For v3 that's gt and challenge; for v4 it's captcha_id and, if you know it, the kind.

  2. 2

    You send them to Takion

    One GET to the /v3 or /v4 endpoint with your key and those params. Optionally a proxy to solve under. No SDK, no widget to embed.

  3. 3

    We solve it server-side

    Takion runs the challenge logic the GeeTest JS would have run, works out the slide offset or the icon picks, and produces the token set the site verifies against.

  4. 4

    You submit the returned fields

    The response carries the full token set for your version. Drop those values into your form or request exactly where the widget would put them, and the check clears.

v3 and v4, decoded

Two generations, two payload shapes. Takion returns the right one for whichever the site runs:

v3 (in)
the captcha's gt and challenge. gt is the site's fixed public ID; challenge is minted per attempt and expires fast.
v3 (out)
geetest_challenge, geetest_validate, and geetest_seccode (the seccode is the validate value with a suffix appended).
v4 (in)
the captcha's captcha_id, plus an optional kind for the challenge type.
v4 (out)
captcha_id, lot_number, pass_token, gen_time, and captcha_output.
kind
which surface to solve: slide puzzle, icon / point-selection grid, or ai / space captcha. Set it to match what the site serves.

Why request-based beats a headless browser here

The browser way to beat GeeTest is a headless Chrome that loads the widget, renders the puzzle image, and fakes a mouse path dragging the slider to the gap. It works until GeeTest tweaks the challenge, and it costs you a page render, a real viewport, and seconds per solve. Scale that to thousands of logins and you're running a Chrome farm just to drag one slider.

Takion never opens the widget. We reverse the challenge and generate the validation payload directly, so a solve is one HTTP round trip that returns a token set instead of a rendered image. No Selenium to babysit, no screenshots, no retries when a container OOMs mid-puzzle. You send the params, you get the fields back, you submit them yourself.

What breaks a GeeTest solve

A GeeTest token is short-lived and tied to the attempt it was minted for. These are what usually turns a valid solve into a rejected one:

  • A stale challenge. The v3 challenge value expires quickly. Read it and solve in the same flow; don't cache it and come back later.
  • The wrong kind on v4. Ask for a slide solve on an icon challenge and the site rejects it. Match the kind to the surface the widget actually shows.
  • Submitting the fields in the wrong place. The token set has to land in your request exactly where the widget posts it. Miss a field or misname it and the verify call fails.
  • Reusing a solved token. One token clears one submission. It's single-use; solve a fresh challenge for each request.
  • An occasional unsolvable challenge. Like any captcha, some challenges just don't solve. That's expected. Load a new one and solve again instead of retrying the dead one.

The one rule that keeps a solve valid

Read the challenge and submit the token in one continuous flow. GeeTest tokens are single-use and short-lived, so a solve is only good for the attempt it was minted against. Solve fresh, submit the exact fields Takion returns, and the check clears.

GeeTest 求解器常见问题

能。Takion 会返回 v3 的令牌集（geetest_challenge、geetest_validate、geetest_seccode）和 v4 的令牌集（captcha_id、lot_number、pass_token、captcha_output）。传入你目标站点所用的版本即可。
滑块拼图、图标／点选网格，以及 AI（空间）验证码。为你目标站点所呈现的挑战设置对应的 kind 即可。
把返回的各字段放进站点用于校验的表单或请求中，就像组件本身会做的那样。文档中给出了各版本下的放置位置。

所有其他防护墙

DataDomePerimeterX / HUMANNuDataCastleAWS CognitoIncapsula / ImpervaAkamai Bot ManagerCloudflarehCaptcha

今天就上线你的 GeeTest 求解方案。

免费试用，之后一个密钥通吃每一堵墙。一小时内完成首次调用。

免费试用定价