OpenClaw · Pay-per-call · No subscription

Crawl the web
with OpenClaw, pay per call.

Enter a URL and fetch the data. Your agents crawl web pages, extract rich metadata, and build datasets — without you managing scraping infrastructure. Billed per call.

  • Single OpenClaw tool — one POST
  • Metadata · clean text · full HTML
  • Per-run budget caps honored
  • Respects robots.txt
AI agent
OpenClaw
Web Fetch
Web Crawling API
robots.txt compliance 100% Every OpenClaw tool call respects robots.txt. Ethical crawling by default — data sourced within the rules the web has set.
Output formats 3 in one call Clean text, full HTML, and structured metadata returned in a single tool response. No follow-up calls needed.
Time to wire it in ~5 min Register one tool in OpenClaw, declare per-call cost, ship. No scraping account, no proxy setup.
What OpenClaw builders ship

Real OpenClaw agents that rely on this tool.

Each pattern below is a single OpenClaw agent with Web Fetch via Locus registered as one tool. The math works because the agent only spends when it actually fetches a page.

OpenClaw SEO agent

Audit and compare pages at scale.

An OpenClaw research agent receives a list of URLs, fetches each page as a tool call, and assembles a metadata audit — titles, descriptions, canonicals, Open Graph tags, heading structure. Per-call billing keeps each audit proportional to the number of pages checked; idle weeks cost zero.

Fetch https://competitor.com/blog/post, return title, meta description, H1–H3s, canonical URL, and Open Graph tags
OpenClaw RAG agent

Fetch web context for any query.

An OpenClaw RAG agent receives a question, identifies relevant URLs, fetches each page as a tool call, and injects clean text into the LLM context window. No proxy management, no rendering pipeline — the tool returns embedding-ready text in the same call that fetches the page.

Fetch https://docs.example.com/api-reference, return clean markdown text ready for embedding
OpenClaw dataset agent

Build LLM training data ethically.

An OpenClaw agent crawls a list of public URLs to build training datasets. robots.txt is respected on every call — data is sourced within the rules the web has set. Clean text output from the tool reduces preprocessing overhead before tokenization.

Fetch https://news.example.com/article, return clean text body respecting robots.txt for dataset ingestion
OpenClaw monitoring agent

Track competitor pages for changes.

A scheduled OpenClaw agent fetches competitor pricing pages, documentation, or news sources on a cron and diffs the structured output. The agent only spends when a fetch fires; OpenClaw budget caps keep cost predictable across high-frequency monitoring schedules.

Fetch https://competitor.com/pricing, extract price table and return structured JSON for change detection
OpenClaw-ready in two minutes

Stop building scrapers into your agent.

Register one tool, connect a wallet, and your OpenClaw agent can fetch any web page on every relevant call. Idle runs cost zero. Scraper migrations cost zero. That's the entire deal.

  • Single OpenClaw tool
  • 100% robots.txt compliance
  • Budget caps honored
  • MCP + pay-per-call native
FAQ

OpenClaw-specific questions.

If something below doesn't cover your case, ping us — we work directly with OpenClaw builders, no SDR funnel.

How does this register as an OpenClaw tool?

+

It's a standard POST endpoint with a typed body. You register it in OpenClaw the same way you'd register any HTTP tool — endpoint, schema, per-call price. OpenClaw uses the price to enforce budget caps and to show the user what each agent run will cost before it runs.

Does OpenClaw need a scraping account or proxy service?

+

No. The gateway manages rendering, retries, and delivery entirely. OpenClaw pays per call from a wallet you connect — no scraping vendor account is created, and there's no proxy credential to rotate or leak in agent logs.

What happens when an OpenClaw run hits its budget cap mid-crawl?

+

The tool returns whatever page data it has resolved so far and OpenClaw stops issuing new fetch calls. The agent treats this as a normal tool boundary and can ask the user for more budget or hand off the partial dataset.

Does the tool respect robots.txt?

+

Yes, always. Every call checks robots.txt before retrieving the page. If crawling is disallowed, the tool returns a clear error. OpenClaw agents can handle this as a normal tool boundary — skip the URL, log the miss, or surface it to the user.

Can OpenClaw call this concurrently across many agents?

+

Yes. Per-key rate limits apply with concurrency-fair queueing — a high-volume agent never starves another agent on the same key.

Does the tool handle JavaScript-rendered pages?

+

Yes. The gateway renders JavaScript before returning content, so single-page apps and dynamically loaded pages are captured in the same tool call.