The infrastructure everybody rebuilds
Yahia Bakour's frustration was specific, and it is the kind that builds companies. He had watched teams repeatedly rebuild — and then get stuck maintaining — the same web data infrastructure: scrapers that break when a site redesigns, crawlers that need babysitting, enrichment pipelines that quietly go stale. Before Context.dev, he led engineering at Sunrun, worked at Amazon, and built and sold two startups. He knew the pattern from the inside: the infrastructure was nobody's core product, but everybody needed it, and it ate engineering time in every company that touched it.
So he started Brand.dev, the company that became Context.dev, with what he calls a simple, stubborn idea: any developer, anywhere in the world, should be able to build software that understands the world it is serving. The opposite of the status quo he described — self-serve, no lock-in, cancel anytime, great data at a reasonable price, behind an API you can call in a couple of minutes. The job, as it clarified over time, was even simpler to state: give agents the internet.
One API for the live web
Context.dev turns the live web into structured data for AI agents and applications. Give it the sources a product cares about and tell it what context is needed; it collects the data, structures it, and keeps it current, without the team building the infrastructure underneath. One REST API covers scraping any URL into clean markdown or HTML, searching the live web, crawling entire sites, extracting brand data like logos, colors, fonts and socials, resolving transaction descriptors, and more — with typed SDKs for TypeScript, Python and Ruby, and an MCP server for agent frameworks.
The company announced itself to the Y Combinator community as a four-person team building the context layer between the internet and every agent or application that needs to understand it, and it found early users among teams that live on web data: Mintlify, Daily.dev, Ferndesk and others were named on its Product Hunt launch, and its own blog later described teams keeping data on more than 20 million companies fresh, powering thousands of integrations, and shipping an onboarding MVP in an hour. The pitch to developers is deliberately boring in the best sense: no brittle scraping infrastructure, no maintenance treadmill, just structured data on demand.
A $9M seed, then 1 billion people
On 19 October 2026, Bakour announced a $9 million seed round led by General Catalyst, with Robinhood Ventures, Orange Collective and other funds participating. In the same announcement he put a number on traction: more than 800 customers. Four days later, Context.dev launched the product this story is about: a People Search API, pitched as 1 billion people, one API, for agents looking for decision makers to sell to, founders to build with, or engineers to hire.
People Search works the way an agent would want it to. Developers describe the people they want in natural language — for example, VPs of sales at solar companies in California — optionally filter by country and years of experience, and get up to 10 profiles per request, with role, employer, location, skills and schools. Each page costs 10 credits, and People Enrich, which adds detail to a known person, costs 20 per match. It adds to a people-enrichment API the company introduced in August 2026, and it sits inside the same credit system as everything else: a free plan with 1,000 credits a month, and paid plans from a $20-a-month Developer tier upward.
That sequencing matters. People Search is not a standalone directory bolted onto a scraping company. It is the same context layer, pointed at a new entity type. The company data, the brand data and the people data all flow through the same API, the same key and the same billing — which is exactly why an agent developer would rather call Context.dev than assemble three vendors and a scraping team.
Coverage is a claim; freshness is the product
There is a useful skepticism to keep in view. A claim of 1 billion profiles is a claim about coverage, not about freshness or accuracy, and Context.dev's own documentation is candid that brand retrieval may reuse cached data up to a documented age. People data goes stale faster than company data — people change jobs, titles and locations — so the value of People Search will depend less on the headline billion and more on how fresh each profile is when an agent acts on it. The company's answer, structurally, is the same as its founding idea: continuous refresh, not one-time extraction.
What Bakour built, in the end, is infrastructure that gets more valuable as agents get more ambitious. A chatbot that answers questions needs web search. An agent that does outreach, hiring research or sales prospecting needs people — described in plain language, returned as structured data, priced per page. Context.dev started by giving developers the web. With People Search, it is giving their agents the people on it. That is a small sentence with a large market behind it, and it is why a four-person team with a $9 million seed could plausibly matter far beyond its size.
The playbook
- Sell the plumbing, not the project — Every team was rebuilding the same scraping, crawling and enrichment plumbing. A single API that collects, structures and refreshes web data turns that undifferentiated work into a call.
- Let the agent describe who it wants — People Search takes a natural-language description — VPs of sales at solar companies in California — and returns structured profiles. The interface is a sentence, not a query language.
- Bundle the context layer — Scrape, search, crawl, brand, enrichment and now people data share one credit system and one key, so each new API makes the whole platform stickier.
- Rename when the job clarifies — Brand.dev became Context.dev when the job got clearer: give agents the internet. A rename that follows a clarified job beats a rebrand that chases a trend.