Cloudflare Turnstile solver library (no headless browser required)

Hacker News Top Tools

Summary

cfsolve 是一个 Rust 库和命令行工具,可在无需无头浏览器的情况下解算 Cloudflare Turnstile 小部件和挑战页面,通过浏览器指纹模拟返回 Turnstile 令牌或 cf_clearance cookie。

No content available
Original Article
View Cached Full Text

Cached at: 10/31/26, 12:15 AM

chris124567/cfsolve

Source: https://github.com/chris124567/cfsolve

cfsolve

cfsolve is a Rust library and command line tool that solves Cloudflare Turnstile widgets and challenge pages without a headless browser. It returns a Turnstile token for widgets or a cf_clearance cookie for challenge pages.

Turnstile widgets can be solved using the page URL, the sitekey, and optionally the page’s HTML (this should not be required in most cases). Challenge pages require a direct fetch or the HTML of the page.

Browser profiles

cfsolve impersonates a specific browser, described by a profile file. One is included: profiles/safari-26-ios.json (Safari on iPhone).

Each profile references recorded responses to the challenge’s browser checks, stored as JSON in a directory. Keep each profile with its directory when you copy. Producing these files requires some manual (or AI) effort though it can be greatly aided by solving challenges in a real web browser with the proxy set to the address of a running capture-proxy, which records the decoded payloads sent to Cloudflare. Currently only Safari and Chrome(ium) profile types are supported.

Building

You need the Rust version pinned in rust-toolchain.toml, plus CMake, a C/C++ compiler, libclang, and Git (to patch and build BoringSSL).

cargo build --release --bin challenge

Command line tool

./target/release/challenge --page 'https://example.com/login' /tmp/solve \
  --solve --fingerprint profiles/safari-26-ios.json

On success it prints the token (or cf_clearance=...) and sends progress to stderr. The directory argument holds all requests and responses for debugging, so it must be empty or not exist yet.

By default it solves the first Turnstile widget on the page, or clears the challenge page. To pick one widget when the page has several, pass its sitekey, with its action and cData when the site sets them. With --sitekey, it fails on a challenge page.

./target/release/challenge --page 'https://example.com/login' /tmp/solve \
  --solve --fingerprint profiles/safari-26-ios.json \
  --sitekey 'SITEKEY' --action 'login' --cdata 'CUSTOM_DATA'

If you only have a sitekey, add --blank-page and set --page to the site’s real URL.

OptionDescription
--fingerprint <file>Browser profile file to use. Required.
--page <url>The page with the widget, or the challenge page.
--solveSolve the widget or challenge. Without this, cfsolve stops after sending the widget’s first payload.
--sitekey <key>Select a widget by sitekey.
--action <value>, --cdata <value>Override the widget’s action and cData. This needs --sitekey.
--blank-pageSkip loading the page. Render the widget on a blank page at the page URL. This needs --solve and --sitekey.
--proxy <url>Send every request through an http://, https://, socks5://, or socks5h:// proxy. Put credentials in the URL, as in socks5h://user:password@host:1080.
-h, --helpShow help.

Library

Add cfsolve as a path dependency and copy its [patch] section, since Cargo only reads patches from the top-level Cargo.toml:

[dependencies]
cfsolve = { path = "/path/to/cfsolve", default-features = false }
tokio = { version = "1", features = ["macros", "rt-multi-thread"] }

[patch.crates-io]
btls-sys = { path = "/path/to/cfsolve/vendor/btls-sys" }
wreq = { path = "/path/to/cfsolve/vendor/wreq" }
http2 = { path = "/path/to/cfsolve/vendor/http2" }

default-features = false turns off the internal tools so only the solver remains. Use the Rust version in rust-toolchain.toml or newer.

use cfsolve::{Browser, Options, PageSource, Turnstile};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let browser = Browser::read("/path/to/cfsolve/profiles/safari-26-ios.json")?;
    let options = Options::default();

    let turnstile = Turnstile {
        url: "https://example.com/login".parse()?,
        page: PageSource::Navigate,
        widget: None,
    };
    let token = browser.solve_turnstile(turnstile, &options).await?;
    println!("{token}");

    let cleared = browser.clear(&"https://example.com/".parse()?, &options).await?;
    println!("cf_clearance={}", cleared.clearance);
    println!("{} cookies; protected page: {}", cleared.cookies.len(), cleared.page.url);
    Ok(())
}

Load a Browser once and share it. Concurrent operations each get their own connections and cookies.

Send the cf_clearance cookie from the same IP address and user agent as the profile, or Cloudflare will reject it.

Turnstile with a sitekey or your own HTML

PageSource::Navigate fetches the page. With only a URL and sitekey, use PageSource::Blank and specify the widget:

use cfsolve::{PageSource, Turnstile, Widget};

let turnstile = Turnstile {
    url: "https://example.com/login".parse()?,
    page: PageSource::Blank,
    widget: Some(Widget {
        sitekey: "SITEKEY".into(),
        action: Some("login".into()),
        cdata: None,
    }),
};
let token = browser.solve_turnstile(turnstile, &options).await?;

Set action and cdata to the values the site uses, or None when the site doesn’t set them. You can also pass a Widget with PageSource::Navigate to pick one widget out of several or to supply one the site adds with JavaScript.

For a page you have already loaded, such as one behind a login, pass its URL and HTML:

PageSource::Html {
    html: std::fs::read_to_string("account.html")?,
    referrer_policy: None, // Or the response's Referrer-Policy header.
}

The widget runs in a cross-site iframe, so you don’t need your login cookies. With Html, widget: None solves the first widget on the page, while Blank always needs a Widget and renders it on a blank page at the site’s URL.

Challenge pages

browser.clear(&url, &options) fetches the page and clears the challenge. The Clearance it returns contains the cf_clearance cookie, all cookies for the page, and the protected page. It fails with Error::NotChallenged when there’s no challenge.

If you fetched the challenge page yourself, pass the response as a FetchedPage to browser.clear_fetched(page, &options). Use the same IP and user agent as the profile.

If you don’t know whether a URL shows a widget or a challenge page, use browser.solve, which covers both cases:

use cfsolve::Outcome;

match browser.solve(&url, &options).await? {
    Outcome::Token(token) => println!("{token}"),
    Outcome::Cleared(cleared) => println!("cf_clearance={}", cleared.clearance),
}

Options

Set a proxy per operation through Options, using HTTP, HTTPS, or SOCKS5:

let options = cfsolve::Options {
    proxy: Some("socks5h://127.0.0.1:1080".into()),
    ..Default::default()
};

Library operations are silent by default. For progress messages and debug files, implement Sink and set options.sink to Some(Arc::new(your_sink)).

Errors

ErrorCauseRetry?
Error::NetworkConnection failed.Yes
Error::RejectedCloudflare rejected the solution.Yes
Error::BlockedThe site returned an error page instead of a challenge.No
Error::ChallengedYou asked for a Turnstile token but got a challenge page.Yes, after clearing it
Error::NotChallengedYou asked for a challenge clearance but there was no challenge.No
Error::ConfigInvalid profile or options.No
Error::UnsupportedAny other failure, such as a new challenge build or a page with no widget.No

License

MIT, except for the patched dependencies in vendor/, which keep their own licenses.

Similar Articles

@vintcessun: An 8-stage vulnerability discovery agent that runs on a Claude subscription, essentially a reproduction of the Cloudflare Project Glasswing paper. Multiple narrow agents + intentional disagreement verification + reachability gating, breaking down 'which piece of code can actually be exploited by an attacker' into 8 precise steps. Recon uses Opus to decompose tasks, H…

X AI KOLs Timeline

Introduces an 8-stage vulnerability discovery agent based on Claude subscription, reproducing the core ideas of the Cloudflare Project Glasswing paper. It uses multiple narrow agents, intentional disagreement verification, and reachability gating to break down exploit analysis into precise steps, without the need for API keys.

byJoey/cfnew

GitHub Trending (daily)

CFnew is a proxy tool based on Cloudflare Workers/Pages, supporting multiple protocols such as VLESS, Trojan, xhttp. It has built-in latency testing, subscription conversion, and a graphical management panel. The v2.9.8 update implements internal subscription conversion, no longer relying on external services, and optimizes transfer performance.

DeCloudflare

Hacker News Top

DeCloudflare is a project hosted on GitLab that provides tools to bypass or interact with Cloudflare services.