Definition
An extract endpoint is a search API method that takes a URL as input and returns the page's content as clean markdown (or structured JSON), replacing the agent-side problem of feeding raw HTML to an LLM and burning ~10x the tokens.
In Depth
Most search APIs in 2024 returned snippets only; agents that needed full-page content had to fetch the URL themselves and convert HTML to markdown. By 2026, extract endpoints became table stakes. Scavio's /extract takes `{url, format: 'markdown'}` and returns `{markdown, title, url}`. The cost is 1 credit per extract ($0.0043 on the $30/mo tier). Common usage: search returns 5 candidates; extract pulls full markdown for the 1-2 most promising. Pair with cache for repeat URLs. Honest constraint: extract works on indexed/public targets; auth-gated or JS-only pages still need a real browser via Browserbase or Stagehand.
Example Usage
Switching from manual HTML fetch + parse to the /extract endpoint dropped the agent's per-task token usage by ~85% and reduced the agent code from ~150 lines (HTTP + Cheerio + markdown converter) to one HTTP call.
Platforms
Extract Endpoint is relevant across the following platforms, all accessible through Scavio's unified API:
Related Terms
HTML Token Cost
HTML token cost is the LLM input cost of feeding raw HTML into a context window versus a cleaner format like markdown; a...
Multi-Platform Search API
A multi-platform search API is a single REST endpoint that returns structured JSON from several public surfaces — Google...
Structured Search Output
Structured search output is the typed JSON returned by a search API — title, snippet, link, position, timestamp — that f...