OpenClaw — autonomous browser agents
Headless browser automation for catalog enrichment, competitor research, and bulk data tasks.
Overview
OpenClaw is K99's headless-browser agent fleet. Where Yen is a conversational interface for shoppers, OpenClaw is a tool for operators: it drives a real Chromium browser to research, scrape, fill forms, and enrich data — at the scale and speed of a small offshore team, but without the overhead.
What OpenClaw agents do
- Catalog enrichment — given a list of SKUs, write SEO descriptions, tag stones, classify by style, and propose collection assignments.
- Competitor monitoring — visit a list of competitor URLs daily and report price, availability, and new-arrival diffs.
- Vendor portal sync — log into supplier portals, download stock CSVs, and reconcile with your inventory.
- Lead research — given a company name, find the buyer's email, LinkedIn, and recent press for the Outbound team to use.
- Compliance checks — confirm every storefront page renders without a console error in 24 languages.
How an OpenClaw run works
- You pick (or write) a playbook — a YAML-defined sequence of goals and tool permissions.
- OpenClaw spins up a sandboxed Chromium with the playbook's starting URL.
- The agent observes the DOM, decides on the next action (click, type, scroll, scrape), and logs every step with a screenshot.
- Whenever it hits a CAPTCHA, payment screen, or login it can't solve, the run pauses and pings a human in
/admin/openclaw/inbox. - On completion, results are written back to your database (products updated, leads enriched, etc.) and a summary is emailed to the operator.
Built-in playbooks
OpenClaw ships with a library of audited playbooks. The most-used ones today:
| Playbook | Typical run | Cost / run |
|---|---|---|
catalog-enrich | ~6s per SKU | $0.002 / SKU |
price-monitor | ~30s per URL | $0.01 / URL |
lead-enrich | ~12s per lead | $0.03 / lead |
storefront-smoke-test | ~3 min per locale | flat $0.50 / run |
Warning
OpenClaw obeys robots.txt and rate-limits to one request per second per origin by default. You can override per-playbook, but the legal and ethical responsibility for scraped data is on the tenant — check the ToS section below.
Writing your own
Custom playbooks live in your tenant under /admin/openclaw/playbooks. They are short YAML files with a goal, an optional start_url, a list of allowed tools (click, type, scrape, screenshot), and a JSON schema describing the expected output. The same playbook can be invoked ad hoc, on a cron, or by a webhook from another K99 surface (Outbound, for example, calls lead-enrich automatically when it imports a CSV).
Terms & safe-use
- OpenClaw is for first-party data, public information, and sites you have an account on.
- Don't use OpenClaw to bypass rate limits, scrape proprietary content for resale, or impersonate users.
- Every run is logged with full screenshots; we will provide them to your counsel if there is ever a dispute.