Introducing Highlights: the context that matters

Build research agents. Cited answers, at any scale.

Ask a research question and get structured findings with source URLs from the Answers API, or run your own agent loop on Search and Scrape. From one company to a million pages.

100 Fast-mode answers free every month · no card required

Who owns the Grand Meridian Hotel, and who manages it?

fast

RESEARCH RUN

  1. Searched the web4 queries
  2. Read pages9 pages
  3. Read company records2 records
  4. Filled your JSONdone

Structured answer with sources

10 credits

POST /v1/web/answers

Research agents built on Context.dev

  • Sixtyfour
  • Bystreet
  • Similarweb
  • OpenTag
  • Gene Inspector
  • Locke
“It’s been very plug and play. We’ve done well over a million scrapes with it at this point and haven’t had any issues.”
Cameron HakeCo-founder, BystreetRead the Bystreet story
scrapes behind Bystreet’s ownership research, without issues
1M+
for Sixtyfour to test production research workloads
< 1 hour

Ask a question. Get cited answers.

Describe the task and the JSON shape you want back. Answers searches the web, reads the pages, and fills in your fields with the URLs it used.

  • SearchFinds the relevant pages on the live web
  • ReadPages and company records, rendered and parsed
  • ReturnYour JSON shape, with source URLs
npm install context.dev

research.ts

import ContextDev from 'context.dev';

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

const research = await client.web.answers({
  task: 'Who owns the Grand Meridian Hotel, and who manages it?',
  mode: 'fast',
  json_format: {
    owner: '',
    operator: '',
    source_urls: [''],
  },
});

console.log(research.json_content);
console.log(research.sources);

What teams research with Context.dev.

Research agents our customers run in production, from hotel ownership to genetics.

  • Ownership and company research

    Research the ownership structure of every hotel, across a million-plus scrapes.

    Bystreet
  • People and GTM research

    Give an AI coworker live, sourced people and company research for CRM work.

    OpenTag
  • Scientific literature

    Connect a research agent to open-access studies and current genetics research.

    Gene Inspector Pro
  • Market intelligence

    Give AI agents the web datapoints they need without building it in house.

    Similarweb
  • Government affairs

    Turn agency pages into research, then watch those sources for changes.

    Locke
  • Research at production scale

    Scraping infrastructure for research agents, tested on production workloads in an hour.

    Sixtyfour

Skip the research loop you’d have to build.

What a homegrown research agent needs before it returns a sourced answer, and what Answers does in one call.

CompareHomegrown agent loopbuild and maintainContext.devone API key
Finding sourcesA search API and your own query planningSearch built into every answer
Reading pagesScrapers, headless browsers, and proxiesPages rendered and parsed for you
Structured outputPrompting and validating JSON yourselfYour json_format, filled in
CitationsTracking URLs through every stepSource URLs returned with each answer
CostLLM tokens plus every API in the loop10 credits in Fast mode, 100 in Ultra

Research agents, answered.

More details in the API documentation.

What is a research agent?

A research agent takes a question, searches the web, reads the relevant pages, and returns findings with the sources it used. With Context.dev you can call the Answers API for a structured, cited answer, or build your own loop on the Search and Scrape APIs.

What’s the difference between Fast and Ultra mode?

Fast mode costs 10 credits per answer and suits quick lookups. Ultra, the default mode, costs 100 credits per answer and researches each question in more depth.

How do research agents avoid hallucinations?

Every Answers response includes the source URLs it read, so you can check each finding or show citations to users. Asking for a source_urls field in your json_format ties each answer to its evidence.

Can I run research across thousands of companies?

Yes. Call Answers in parallel within your plan’s concurrency, or scrape source pages in bulk with Batches, up to 25,000 URLs per job, and run your own agent over the results. Bystreet has run well over a million scrapes for its hotel research.

Run your first research task.

1,000 free credits every month, enough for 100 answers in Fast mode. No credit card required.