DeepLook
LanguagesEnglish

Languages

EnglishEnglishPortuguêsPortuguese · Brasil
Web scraping · Open source

Scrape pages.
Extract content.

Render public pages in Chrome and extract visible text and metadata. Feed your agents, datasets, and crawling pipelines through one API.

Read the quickstart

Engine, CLI, MCP and API under Apache-2.0. Run on your infrastructure.

LIVE API

Try DeepLook live.

Checking connection…

No account. No API key.

Extraction mode

Extract content from one public page.

Examples:

Public pages only. DeepLook visits the URL you submit.

Ready to run—
Run a request to see the content returned by the browser. No response has been preloaded.

Response displayed as text. No page HTML is executed on this website.

Request & demo limits

Loading demo limits…

API REQUEST
POST /v1/open
Content-Type: application/json

{
  "url": "https://example.com",
  "timeoutMs": 20000
}

The request runs through the demo server. API credentials are never exposed in your browser.

In Crawl mode, the demo discovers links on the same website and extracts up to 3 pages, including your starting URL. The API still receives one URL per call.

See the crawling recipe →

Use URLs without credentials or confidential information.

Public URLChromeText + metadataYour pipeline
01 / Scrape · Extract · Crawl

From public pages
to your content pipeline.

Scrape a page, extract its content, or use DeepLook as the extraction layer in your crawler.

01

Scrape browser-rendered pages.

Open a public URL in a real browser, render page JavaScript, and retrieve its visible text.

SCRAPING / GOOGLE CHROME

02

Extract usable content.

Get text, title, final URL, and HTTP status in JSON. Use the content in your agents, indexes, and datasets.

EXTRACTION / TEXT & METADATA

03

Build crawling workflows.

Supply URLs from your queue and use DeepLook to extract each page. Your crawler controls discovery, deduplication, and traversal.

YOUR QUEUE / DEEPLOOK EXTRACTION

02 / The scraping API

One URL in.
Page content out.

Scrape a public URL with POST /v1/open. Extract text and metadata as JSON; challenges, empty pages, and timeouts return explicit errors.

Read the API and error reference
See a multi-page extraction recipe 
curl "$DEEPLOOK_BASE_URL/v1/open" \
  -H "Authorization: Bearer $DEEPLOOK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com","timeoutMs":30000}'
YOUR INSTANCE BASE URL AND KEY · RUN ON YOUR SERVER

This example follows the current API. A public production endpoint is not available yet.

03 / Choose where to run

Local by choice.
API with your key.

Use the local engine or connect your agent to the hosted API. Engine, CLI, MCP and HTTP adapter remain available under Apache-2.0.

Local engine

APACHE-2.0

Run the browser on your infrastructure and connect extraction to your scraper or crawler.

  • Engine, CLI and stdio MCP
  • Node.js + Google Chrome
  • Control over your local execution
  • Open source under Apache-2.0
Build from source

Clone the GitHub repository and build locally.

Hosted API

VERIFIED ACCOUNT

Verify your email, create a key and connect your agent at api.getdeeplook.dev.

  • POST /v1/open with a Bearer token
  • Individual, revocable keys
  • Per-account quotas and concurrency
  • Private account dashboard
Get an API key

Account creation and the newsletter are independent.

Stdio MCP is available. Connect your client to the CLI with deeplook mcp and use the deeplook_open tool. The MCP adapter does not expose an HTTP endpoint.

Configure MCP →
04 / Before you integrate

What to expect
from DeepLook.

Does DeepLook crawl a whole website?

The playground demonstrates a crawl of up to 3 pages, discovering links on the same website. The current endpoint extracts one URL per request. For multi-page crawling, your application supplies the URL queue and handles link discovery, deduplication, limits, and retries. DeepLook scrapes and extracts each page; there is no native crawling endpoint yet.

What content is extracted?

Visible text rendered by Chrome and page metadata, including the title, final URL, and HTTP status. The current API does not provide custom field extraction, Markdown, or a ready-made dataset.

Does DeepLook search or run full research tasks?

The engine opens a public URL and returns its visible text. Your agent or application handles search, source selection, and writing answers. Nymetric is a separate research product.

Does it work on every site?

Access to every page is not guaranteed. Bot challenges, access barriers, empty content, and navigation failures return explicit errors. DeepLook does not solve captchas or promise to bypass blocks.

Can I install the CLI or connect through MCP yet?

Yes. Clone DeepLook on GitHub, install dependencies and build. The CLI supports local reading and deeplook mcp over stdio. See the quickstart and MCP configuration in the docs.

How can I use the API?

Create an account with verified email to generate a key and use the hosted API at api.getdeeplook.dev. DeepLook-API is also open source under Apache-2.0 for your own infrastructure.

What happens when a request is interrupted?

The API cancels browser work when the client disconnects or the deadline expires, and waits for process cleanup before freeing the concurrency slot.

Your next content pipeline
starts with a URL.

Create your key and connect your agent to your next scraping workflow.

Get an API key