Stealthrekt beta
Are your AI agents stuck on CAPTCHA?
Give your agents a CAPTCHA-capable browser that detects challenges, runs supported solvers, resumes the task, and saves proof JSON, screenshots, and replay media for every attempt.
Proof, not vibes. Replay the proof · View proof JSON · Replay GIF
CAPTCHA solving, with the browser stack to back it up
CAPTCHA detection
Detect challenge surfaces inside real browser tasks instead of leaving agents stuck at the wall.
Supported solver pipeline
Run Stealthrekt-owned fixture/solver interfaces, capture the resolution method, and continue the browser flow after the challenge clears.
Replayable proof
Return proof JSON, screenshots, replay GIFs, MP4 clips, final URL, confirmation text, and solver evidence.
Fingerprint hardening
Chrome flags, injected patches, user-agent rotation, navigator controls, canvas/WebGL/audio noise, and seeded profiles help the solver reach the challenge.
Fast browser starts
Optimized launcher plus pre-warmed context pools keep CAPTCHA attempts fast enough for agent workflows.
Proxy-ready sessions
Per-session proxy configuration, credential redaction, health scoring, and rotation/failover hooks support production-grade solving.
Why the CAPTCHA solver can hold up
Solving is the product; stealth runtime is the moat
Stealthrekt packages the browser infrastructure around CAPTCHA solving: fast launch, durable contexts, fingerprint controls, proxy diagnostics, challenge evidence, and release gates that keep claims tied to artifacts.
CAPTCHA proof flow
Verified locally · `captcha_form` produces proof JSON, screenshot evidence, and replay media.
Pre-warmed contexts
Implemented · `ContextPool` tracks warm hits, cold starts, acquisition latency, release latency, and pool health.
Launch metrics
Implemented · `ChromeLauncher` records Playwright startup, browser launch, context creation, page creation, and total launch time.
Fingerprint profiles
Implemented · seeded profiles cover UA, platform, hardware concurrency, device memory, screen, canvas, WebGL, audio, and fonts.
Agent CAPTCHA benchmark prompt
Copy a concrete CAPTCHA-solving task
Give another agent a measurable CAPTCHA flow instead of a vague product claim.
Use Stealthrekt to run `stealthrekt run captcha_form --json --run-dir runs/agent-proof`, then report whether CAPTCHA detection fired, which supported solver path resolved it, whether the browser task resumed, and where proof JSON, screenshots, final URL, launch metrics, proxy readiness, and replay artifacts were saved.
First run
pip install stealthrekt playwright install chromium stealthrekt doctor stealthrekt run captcha_form --json --run-dir runs/demo