Steel API
Steel API on looot as of 2026-09-28: 3 jobs through 3 endpoints, how you reach it, what it costs, and how to run it.
Steel API
What it’s good for
Steel API is in Web data. Reach for it to scrape a web page into markdown; take a screenshot of a web page; get pdf.
As of 2026-09-28 (catalog revision 85280) it serves 3 jobs through 3 endpoints.
Jobs
| Job id | What it does | Endpoints |
|---|---|---|
pdf.get |
Get pdf | 1 |
web.scrape.markdown |
Scrape a web page into markdown | 1 |
web.screenshot |
Take a screenshot of a web page | 1 |
Access
- Runs on looot’s key for 3 endpoints: nothing to set up, the run spends your looot balance. You can connect your own key instead; see Bring your own key.
Price basis
3 per call (from $0.005). A call that fails at the provider costs $0; see What’s free.
Free first in fallback
No. Every endpoint bills on each answer, found or not, so a fallback walk pays for a miss here. See Free first.
Run it
A job run takes shared input names, not Steel API’s own field names; looot maps them for each provider. For job:web.scrape.markdown, url is taken by 4 of its 4 providers. See Job inputs.
Example input for looot run or the MCP run tool, with placeholder values:
looot run job:web.scrape.markdown --input '{"url":"https://example.com"}'{
"endpointId": "job:web.scrape.markdown",
"input": {
"url": "https://example.com"
},
"idempotencyKey": "<new unique key>"
}A job run picks among all the job’s providers. To run Steel API itself, name one of its endpoints (MCP search with {"filters": {"provider": "steel"}} lists them), for example steel-scrape: looot inspect steel-scrape prints its exact input fields and price.
In the live catalog
- Browse: looot.ai/public-catalog/steel-pdf, or search looot.ai/public-catalog for Steel API.
- Every endpoint, no account needed:
curl "https://api.looot.ai/v1/public-catalog?provider=steel"