Deep research API. Get answers you can use.
Ask a research question and get a structured answer in your JSON shape, with the sources behind it.
Powering agents and products with the world's data.
Ask once. Get answers you can cite.
Describe what you need to know and the shape you need it in. Answers reads the web and returns findings with their sources.
Start with a question
Describe the research task instead of assembling a set of pages by hand.
Choose the structure
Define the JSON shape your product needs, so the result can move straight into your workflow.
Keep the sources
Use the returned page URLs to show your users where the answer came from.
Your question. Your output shape.
Send a task and an example JSON structure. The response includes the structured answer and the URLs used as evidence.
npm install context.devimport ContextDev from 'context.dev';
const client = new ContextDev({ apiKey: 'YOUR_API_KEY' });
const answer = await client.web.answers({
task: 'Find the main product features on linear.app.',
mode: 'fast',
json_format: {
features: [''],
},
});
console.log(answer.json_content);
console.log(answer.sources);Give your agents a research capability.
Move from a question to sourced findings that your application can work with.
Run deep research on demand
Resolve a domain to a typed brand profile, then research the company in your JSON shape, with sources.
Enrich any entity your agent sees
Connect company references and person identifiers to brand profiles and people data.
A few things to know.
More details in the API documentation.
What is a deep research API?
A deep research API takes a question, searches and reads web sources for you, and returns a synthesized answer. Context.dev Answers returns that answer in the JSON shape you define, together with the source URLs it used.
What is the difference between Fast and Ultra?
Fast mode costs 10 credits per answer. Ultra, the default mode, costs 100 credits per answer. Choose the mode that matches how much research the task needs.
How are the sources returned?
The response includes a sources array of URLs that supplied search results or readable page content. Keep those URLs with the answer so your users can inspect its evidence.
Can I choose the answer format?
Yes. Set json_format to the structure your application expects and read the result from json_content.
How is Answers different from page extraction?
Answers begins with a research task and gathers web evidence. Use extraction when you already have a particular page and want specific fields from it.
What if research stops before it is finished?
The response can include a partial flag when the deadline ends research. Your application can distinguish that result from a completed research run.
Put research inside your product.
Ask the web a question and get structured, cited answers. The same API key works for scraping, search, and monitors.