Use case
One scraping endpoint
your agent can trust.
Agents can't debug a CAPTCHA mid-task. gottem gives them a single HTTP endpoint that routes every fetch across 14+ scraping vendors, escalates automatically when a site fights back, and returns the same JSON shape no matter which vendor served the page, so the agent's tooling never changes.
Built to be called by agents
Every call is one HTTP request. No SDK, no session state. The response carries a routing receipt (provider, route, tier, cost, content quality) so an agent can reason about what it got. And the whole integration ships as a paste-in skill: SKILL.md describes when to invoke gottem, llms.txt indexes the full surface for LLM-native discovery.
Failures are handled below the waterline
When a vendor returns a 403, a CAPTCHA, or a Cloudflare wall, gottem retries through a different vendor inside the same request. The agent sees one response, the one that worked. Failed attempts land in the ledger at the $0.000001 minimum, effectively free.
Safe by default
Shield, the malicious-host firewall, blocks known phishing, malware, and scam domains before any vendor is called or charged, so an agent following untrusted links can't be steered into fetching something hostile on your bill.
Start with a curl.
curl -sS "$GOTTEM_BASE_URL/scrape" \
-H "Authorization: Bearer $GOTTEM_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://example.com"}'Questions.
- How does an agent discover gottem's capabilities?
- Two paste-in files: gottem.dev/SKILL.md is a ready-made agent skill describing when and how to call the API, and gottem.dev/llms.txt is the machine-readable index of the whole surface.
- What happens when a fetch fails mid-task?
- gottem retries through another vendor inside the same request, and the agent receives the response that worked. A failed attempt bills at the $0.000001 minimum, effectively free.