Context.dev Review – YC S26 Web Data API for AI Agents

TL;DR

TL;DR: Context.dev is a unified REST API for scraping web pages to markdown, crawling entire sites, extracting structured data via Zod schemas, capturing screenshots, and identifying brands — all through a single authenticated endpoint.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: context.dev ← MUST visit and verify
• Documentation: docs.context.dev ← MUST read
• HN launch thread: news.ycombinator.com/item?id=48847562
• Source last checked: 2026-07-14

What Is Context.dev?

Context.dev is a web data extraction API launched at YC Summer 2026 (S26). It provides a single REST API that handles multiple web data needs: converting pages to clean markdown, crawling entire websites, extracting structured JSON data via user-defined Zod schemas, generating screenshots, and identifying brands from domain or transaction strings.

The core value proposition is unification — instead of stitching together a scraping library, a headless browser, a screenshot service, and a brand lookup API, you call one endpoint with your API key and get back typed, structured data.

Setup Workflow

Step 1: Get an API Key

Sign up at context.dev to receive your CONTEXT_DEV_API_KEY. The free tier includes a generous allocation to get started.

Step 2: Install the SDK

npm install context.dev
# or
pip install context-dev

Step 3: Scrape a Page to Markdown

import ContextDev from 'context.dev';

const client = new ContextDev({ apiKey: process.env['CONTEXT_DEV_API_KEY'] });

const { markdown } = await client.web.webScrapeMd({
  url: 'https://news.ycombinator.com'
});

console.log(markdown);

Step 4: Crawl an Entire Site

const { results, metadata } = await client.web.webCrawlMd({
  url: 'https://docs.stripe.com',
  maxPages: 50,
  maxDepth: 3,
  useMainContentOnly: true
});

console.log(`Crawled ${results.length} pages`);

Step 5: Extract Structured Data with Zod

import { z } from 'zod';

const schema = z.object({
  company_name: z.string(),
  pricing_tiers: z.array(z.object({
    tier_name: z.string(),
    tier_price: z.number(),
    tier_currency: z.string(),
    tier_billing_model: z.enum(['monthly', 'yearly', 'one_time', 'usage_based']),
  })),
});

const result = await client.web.extract({
  url: 'https://www.context.dev',
  schema: schema.toJSONSchema(),
});

console.log(result.data.pricing_tiers);

Deeper Analysis

API Design

Context.dev’s SDK is available in TypeScript/JavaScript and Python. All endpoints return typed objects — no raw HTML parsing required on the client side. Authentication is API key via environment variable, consistent with standard developer tool practice.

Feature Breakdown

• Web scraping (webScrapeMd) — Renders JavaScript-heavy pages via a headless engine, returns clean markdown. Handles maxPages and maxDepth for large sites.
• Structured extraction (web.extract) — Takes a Zod schema, converts it to JSON Schema server-side, and returns typed data matching your schema shape. This is the most differentiated feature — it eliminates the fragile regex/XPath scraping pattern.
• Screenshots (web.screenshot) — Returns a hosted CDN screenshot URL in viewport or full-page mode.
• Brand intelligence (brand.retrieve) — Identifies a company by domain, name, email, ticker, or direct URL. Useful for building brand databases or enriching user profiles.
• Styleguide extraction (web.extractStyleguide) — Returns color palette, typography, and element inventory for any domain.
• Product extraction (ai.extractProducts) — Scrapes product listings or individual product pages, returning structured price, description, and feature data.

Free Tier

The free tier provides enough API credits to evaluate the full feature set. Paid plans scale by credits consumed per month.

Practical Evaluation Checklist

• Clean markdown output from JavaScript-rendered pages
• Zod schema validation works for custom extraction
• Crawl respects maxPages and maxDepth limits
• Screenshots render correctly in viewport and full-page modes
• Brand lookup returns data for well-known domains
• SDK is well-typed and error messages are actionable

Security Notes

API keys are passed via environment variable — never hardcode them in source code or client-side applications. Context.dev’s serverside processing means raw HTML never touches your infrastructure, reducing exposure to malicious payloads from scraped pages.

FAQ

Q: Does this replace Puppeteer or Playwright? A: For data extraction, yes. Context.dev handles JavaScript rendering server-side. For interactive browser automation (filling forms, clicking elements, testing), you still need Puppeteer or Playwright.

Q: How does it handle rate limiting or large sites? A: The crawl endpoint respects maxPages and maxDepth to prevent runaway crawling. For production-scale crawling, paid plans offer higher throughput.

Q: Is there a Python SDK? A: Yes. The Python SDK covers all the same endpoints as the JavaScript/TypeScript version.

Q: Can I extract data from pages that require login? A: No. Context.dev scrapes publicly accessible pages only. Authenticated content is not supported.

Conclusion

Context.dev solves the fragmented web scraping stack problem by providing a single API that scrapes, crawls, extracts structured data, screenshots, and identifies brands. The Zod schema extraction feature is the standout — it makes fragile XPath/regex scraping obsolete. If your AI pipeline needs reliable web data, this is worth evaluating on the free tier first.

Source and Accuracy Notes

• Project page: context.dev
• Documentation: docs.context.dev
• HN launch thread: news.ycombinator.com/item?id=48847562
• License: Proprietary (API service, no open-source license)
• Source last checked: 2026-07-14