Analysis updated 2026-05-18
Automate a repetitive browser task once and have it run forever unattended.
Schedule 24/7 browser-based jobs like data checks or account tasks.
Build a personal library of learned browser automation skills.
| sediman-agent/openskynet | akmessi/vex | fredantb/spec-driven-development | |
|---|---|---|---|
| Stars | 36 | 36 | 36 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | easy |
| Complexity | 3/5 | 3/5 | 2/5 |
| Audience | developer | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an LLM API key and installing a real browser via Playwright/Chromium.
OpenSkynet describes itself as an AI browser employee that carries out repetitive tasks around the clock. The idea is that a person shows the tool how to do something in a browser once, and it can then repeat that task automatically going forward, including on a schedule. It is installed as a command line tool, either directly from GitHub with a tool called uv or by building it from source. Once installed, a user runs an init command to set an API key, then can either give it a single instruction to run right away or open an interactive chat session. There is also a separate terminal interface built in Rust that lets a user pick an AI provider and model, view and edit the agent's memory, list learned skills, and manage scheduled jobs. Under the hood it uses a real browser through Playwright and Chromium rather than simple scraping, and pairs that with an AI model to decide what to do on a page. The README highlights several features compared to plain browser automation tools, scrapers, and traditional RPA software: learning a task by watching it done once, automatically adjusting itself when a page's layout changes, saving reusable skills after each task, running jobs on a schedule using cron style timing, remembering user preferences across sessions, and splitting complex tasks across multiple subagents that work in parallel. It also mentions a hub where users can browse and install several hundred community made skills. The project is self hosted, meaning it runs on the user's own machine rather than through a cloud service, and it is aimed at people who want to automate repetitive browser based work without writing custom scraping code for every site. It is released under the Business Source License 1.1, which is not a fully open license and places restrictions on some uses.
A self-hosted AI agent that automates repetitive browser tasks by learning from a single demonstration and running them on a schedule.
Mainly Python. The stack also includes Python, Playwright, Rust.
A source-available license (BSL 1.1) that is not fully open source and restricts some commercial uses, check the license terms before using it in a business.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.