- Web Unlocker — stateless read and extraction. Best when you want page content quickly without maintaining a browser session.
- Cloud Browser — stateful interaction. Best when you need navigation, clicks, typing, screenshots, or multi-step flows that persist across commands.
Install
Credentials
Two different Gologin credentials power the two modes:Config file
Store credentials in~/.gologin-web-access/config.json:
config init. It isn’t interactive: pass the keys as flags, or it takes them from the environment variables above. Then it checks the keys against the Gologin APIs:
--no-validate to skip the key check. If the config file already exists, config init leaves it as is unless you add --force.
Quick start — read a page
Quick start — interact with a site
Command reference — scraping / read
These commands use Web Unlocker. They are stateless — no browser session is maintained.Single page
Common options:
--format text|markdown|html, --source auto|unlocker|browser, --fallback none|browser
Batch operations
Batch options:
--format, --source, --only-main-content, --retry <n>, --backoff-ms <ms>, --summary, --output <path>, --strict
By default batch-scrape returns exit 0 on partial success. Add --strict to fail on any error.
Search and discovery
Options:
--limit <n>, --max-depth <n>, --concurrency <n>, --only-main-content
Async crawl jobs
Change tracking
Runbooks
Command reference — browser / interact
These commands use Cloud Browser. They maintain a stateful session through a local daemon.Session management
Options:
--profile <id> to use a specific Gologin profile.
Page reading
Interaction
Navigation
Tabs
Cookies and storage
Advanced
Use cases and examples
Scrape a competitor’s pricing page
nextActionHint suggests switching to browser mode.
Build a research corpus from a docs site
Monitor pages for changes
Extract structured data with a schema
Define what you want to extract. The schema maps each output field to a CSS selector:selector— CSS selectortype—text(default),html,attr, orexistsattribute— attribute name whentypeisattrall—truereturns every match as a listfields— a nested schema applied inside each match
batch-extract.
Search the web and scrape results
Log into a site and interact
When scraping isn’t enough — you need to click, type, navigate:Parse local documents
Not just web pages — parse PDFs, Word docs, Excel files:Automate multi-step workflows with runbooks
Define a sequence of steps in JSON and replay them:Source routing
Commands that support--source can route through different backends:
Structured output
scrape-json returns rich metadata including:
title,description,headings,headingsByLevel,links- Outcome classification:
ok,empty,incomplete,authwall,challenge,blocked,cookie_wall nextActionHint— suggested next step when the page is not fully accessible
search returns requestedLimit, returnedCount, warnings, cacheTtlMs, and per-result position.
Related
- Web Unlocker SDK — TypeScript SDK for programmatic use
- What is Web Unlocker — product overview
- Agent Browser CLI — browser-only CLI for AI agents
- Local runtime — Agent Browser CLI with local Orbita profiles
- AI Skills — plug-and-play skills for Claude Code
npm package
gologin-web-access
Source code
GologinLabs/gologin-web-access