glidercli

Browser automation CLI with autonomous loop execution.

About

What	Control a Chromium-based browser from the terminal via CDP; run YAML tasks; loop until done (Ralph Wiggum pattern).
CDP	Chrome DevTools Protocol via relay + browser extension
Tasks	Declarative steps: goto, click, explore, eval, screenshot
Loops	Run until completion marker or max iterations / timeout
Safety	Max iterations, timeout, backoff

---

Install

Step	Action
1. CLI	npm i -g glidercli
2. Extension	Install Glider from Chrome Web Store (required; bridges relay ↔ tab).
3. Daemon	glider install then glider connect
4. (Optional) Browser	Default: Chrome. For Arc/Edge/Brave: BROWSERS.md or export GLIDER_BROWSER=Arc

Requirements

Requirement	Minimum
Node	18+
Browser	Chromium-based (Chrome, Arc, Edge, Brave, Opera, Vivaldi, DuckDuckGo). No Firefox/Safari → BROWSERS.md

---

Usage

glider connect
glider status
glider goto "https://reddit.com"
glider eval "document.title"
glider run task.yaml
glider loop task.yaml -n 50

Daemon	Logs
glider install / glider uninstall	~/.glider/daemon.log

---

Task files

name: "Reddit"
steps:
  - goto: "https://reddit.com"
  - wait: 2
  - eval: "document.title"
  - screenshot: "/tmp/out.png"

---

Commands

Command	Description
glider install	Install daemon (relay at login)
glider uninstall	Remove daemon
glider connect	Connect to browser
glider status	Server / extension / tabs
glider browser	Show browser config (name, path)
glider goto <url>	Navigate
glider eval <js>	Run JS in page
glider click <sel>	Click element
glider type <sel> <text>	Type into input
glider screenshot [path]	Capture page
glider run <file>	Run YAML task
glider loop <file> [-n N]	Loop until done or limit

Full list: glider --help.

---

Docs

Doc	Contents
This README	Install, usage, commands
docs/BROWSERS.md	Which browsers work; browser name/path (env or ~/.glider/config/browser.json)
config/browser.json.example	Example browser config

---

Roadmap

Status	Area
Done	CDP relay + extension, YAML tasks, loop, daemon, multi-tab
Todo	Linux / Windows, headless, task chaining, scraping templates

---

Contact