Introducing Highlights: the context that matters

Feed your RAG pipeline. Fresh web data, ready to embed.

Turn docs, help centers, and whole websites into clean Markdown for your RAG pipeline, then keep the index current with monitors. No HTML parsers or cleanup scripts to maintain.

1,000 free credits a month · no card required

Sources

crawl

  • /billing/payment
  • /api/keys
  • /guides/sso
  • /changelog

help.acme.com · 312 pages

Markdown chunks

main content

  • ## Payment methods212 tok
  • ## Rotating API keys168 tok
  • ## v4.2 release notesnew

Your vector index

your stack

4,812

chunks embedded

monitor: /changelog changed

1 page re-embedded · 2m ago

crawl → markdown → your embeddings

Teams feeding RAG with Context.dev

  • SiteGPT
  • Hyperspell
  • Agently
  • Sourcely
  • Notra
  • Polished.ad
“I installed the SDK, swapped the scraper service behind our Brain, and deleted the polling code the same day.”
Omar GhandourFounder, AgentlyRead the Agently story
for SiteGPT to move its chatbot knowledge bases
< 1 day
for Hyperspell’s web content and company data
1 API

Ingest, clean, and keep it fresh.

Crawl a site into Markdown with only the main content, chunk and embed it with your own stack, and let a monitor tell you when a source page changes.

  • IngestCrawl a site, map it, or scrape a URL list
  • CleanMain-content Markdown, no nav or footers
  • RefreshMonitors flag the pages that changed
npm install context.dev

ingest.ts

import ContextDev from 'context.dev';

const client = new ContextDev({ apiKey: 'YOUR_API_KEY' });

const crawl = await client.web.webCrawlMd({
  url: 'https://help.acme.com',
  maxPages: 500,
  useMainContentOnly: true,
});

for (const page of crawl.results) {
  // Chunk and embed with your own stack, keyed by the source URL
  console.log(page.metadata.url, page.markdown.length);
}

What teams build on fresh web data.

RAG pipelines our customers run on Context.dev, from support bots to company brains.

  • Support chatbots

    Train each customer’s bot on their own docs and help center, as clean Markdown.

    SiteGPT
  • Company brains

    Combine public website content with Slack, email, and Drive in one source of truth.

    Hyperspell
  • Knowledge at signup

    Read a new customer’s website during signup so the agent knows the business.

    Agently
  • Research corpora

    Crawl journals and PDFs into a corpus a research product can cite.

    Sourcely
  • Brand knowledge bases

    Feed ad generation with crawl-ready Markdown about each brand.

    Polished.ad
  • Chatbots with live search

    Pair a knowledge base with web search and fetch for what it doesn’t know.

    Notra

Skip the ingestion code you’d have to babysit.

What a homegrown RAG ingestion pipeline needs before the first embedding, and what Context.dev handles.

CompareHomegrown ingestionbuild and maintainContext.devone API key
Getting the pagesA crawler, headless browsers, and proxiesCrawl, map, or scrape through one API
Clean textPer-site parsers to strip nav and footersMain-content Markdown on every page
PDFsA separate document parsing pipelineWeb pages and PDFs through the same API
FreshnessRe-crawl everything on a scheduleMonitors flag only the pages that changed
CostServers, proxies, and engineering time1 credit per page, failed pages are free

RAG pipelines, answered.

More details in the API documentation.

What is a RAG pipeline?

A RAG (retrieval-augmented generation) pipeline ingests documents, splits them into chunks, embeds them, and retrieves the most relevant chunks to ground an LLM’s answer. Context.dev handles ingestion for web content: it crawls sites and returns clean Markdown ready to chunk and embed.

How do I get website data into a RAG pipeline?

Crawl the site with Context.dev to get main-content Markdown for every page. Split each page into chunks, embed them, and store the vectors with the page URL so answers can cite their source.

How do I keep a RAG index up to date?

Create monitors for the pages or sitemaps you index. When a run detects a change, the webhook tells you which page changed, so you re-embed that page instead of re-crawling everything.

Can I ingest PDFs as well as web pages?

Yes. Scrape and Crawl turn PDFs into Markdown alongside web pages, and the Parse API converts PDF and Office files you upload.

Feed your pipeline today.

1,000 free credits every month, enough for 1,000 pages of Markdown. No credit card required.