Content extraction endpoint

Web Content Extraction API

Turn a public webpage into content your application can read and store. The endpoint removes much of the page chrome that makes raw HTML hard to use, while keeping fetch evidence beside the result.

Get an API key Read documentation

Web pages passing through a processing grid and becoming structured data cards

Endpoint

POST /v1/extract

A working request

Set your key once in a server-side environment variable, then send a JSON request.

cURL

curl --request POST \
  --url "https://api.tooltrace.io/v1/extract" \
  --header "X-ToolTrace-Key: ${TOOLTRACE_KEY}" \
  --header "Content-Type: application/json" \
  --data '{
  "url": "https://example.com",
  "render": "auto",
  "mode": "structured",
  "include": [
    "markdown"
  ]
}'

Response scope

What the API returns

The response is structured for software, with source evidence kept beside the extracted or inspected data.

Clean Markdown and readable text
Structured page sections
Optional raw HTML
Requested and final URL evidence
Render method, timing, and warnings

Practical uses

AI and RAG preparation

Prepare focused page content before it enters a prompt, search index, or retrieval pipeline.

Research collection

Save readable source material with the final URL, status, and extraction evidence attached.

Publishing workflows

Compare live page content with expected releases or transform approved sources into another format.

Implementation guidance

Use the result with context.

Choose rendering with care

Static retrieval is faster and works for pages that include useful content in the initial HTML. Browser rendering is for pages that depend on JavaScript before their main content appears.

Use automatic rendering when targets vary. If you already know a site is server rendered, choose static retrieval to keep the request simpler and faster.

Keep source evidence

Store the requested URL, final URL, status, render method, warnings, and request ID with extracted content. These fields help an agent or reviewer judge whether the result is complete and current.

Authentication, limits, and errors

Send your key through the X-ToolTrace-Key header from server-side code. Read credit and rate-limit headers after every request. Validation, blocked destination, authentication, plan, and upstream fetch errors use structured responses that applications should handle by code rather than message text.

Use a dedicated key for each environment. Review current fields in the interactive API reference and compare usage limits on the pricing page.