An undetectable Playwright MCP server and browser agent for stealth automation.
Run undetectable AI web browser agents via MCP or a dedicated local web UI.
This tool provides an anti-detect stealth Firefox browser agent designed to bypass captchas and anti-bot systems during automated web tasks. Compliance and audit professionals can use it to securely gather data from dynamic public web sources that typically block standard scraping tools.
It operates as a Model Context Protocol server or standalone web UI that drives browser interactions using mouse movements and keyboard presses just like a human. It supports custom proxies, persistent profiles, and specific browser seeds to maintain consistent browser identities.
It suits compliance officers and auditors needing to inspect heavily protected web portals or verify public-facing disclosures without triggering security blocks. It is not suitable for users who lack technical capability to configure local runtimes, environment variables, or AI model API keys.
Navigates pages using anti-detect stealth Firefox to avoid captchas and blocks.
Functions either as an MCP server for AI assistants or via a standalone local web UI.
Maintains login states and cookies across restarts using dedicated profile directories.
Supports HTTP and SOCKS5 proxies that automatically adjust timezone, locale, and egress.
Free under the MIT license, with no license fees or software costs. Running it requires an external API key from OpenRouter, which incurs usage costs based on the model chosen.
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
No data is sent to the author; sessions, screenshots, and profiles remain strictly on your local machine.
You can specify any OpenRouter model id via flag or environment variable, defaulting to z-ai/glm-5.3-flash.
It maps seeds to fingerprints, preferences, proxies, and geolocations using invisible_core components.
See Zekai first in Google