Former Twitter CEO Parag Agrawal has launched Parallel Web Systems, raising $30 million to develop the Deep Research API. This tool acts like a turbocharged research assistant, letting AI agents browse the web in real-time and gather fresh information with proper citations. The system uses eight specialized engines to handle complex research tasks, already powering millions of daily searches for legal teams, startups, and coding projects. There’s much more to uncover about this game-changing technology.
Former Twitter CEO Parag Agrawal is stepping back into the spotlight with something that might just change how AI agents do their homework. His new company, Parallel Web Systems Inc., just launched the Deep Research API, and honestly, it’s pretty impressive stuff.
Agrawal’s comeback: a research API that could revolutionize how AI agents gather and process real-time information.
Think of it this way: most AI models are like really smart students who memorized everything from textbooks but can’t google anything during the test. Agrawal’s API is different. It lets AI agents actually browse the web in real-time, gathering fresh information and citing sources like a proper research assistant. Similar to how Snowflake Intelligence enables natural language interactions with data, this API simplifies complex research workflows.
The Palo Alto-based startup raised $30 million from heavy hitters like Khosla Ventures and Index Ventures. With only 25 employees, that’s serious funding per person. The team includes talent poached from Google, Airbnb, and Waymo, so they’re not messing around.
Here’s where it gets interesting: the API claims to outperform humans and even GPT-5 on research benchmarks. Eight specialized engines handle different tasks, from long-form synthesis to cross-disciplinary analysis. It’s like having a research team that never sleeps, never gets coffee breaks, and actually cites their sources properly.
The real magic happens in practical applications. Legal teams use it for case research, startups deploy it for market intelligence, and coding agents leverage it for debugging. Instead of relying on outdated training data, these AI agents can pull live information from across the web.
Currently, the system powers millions of research tasks daily for enterprises and ambitious startups. Whether you’re building autonomous agents or need complex decision-support systems, this API promises to embed serious research capabilities directly into your applications.
The timing feels right, too. As AI agents become more sophisticated, they need better ways to gather and analyze current information. Static datasets won’t cut it anymore when you’re making real-world decisions. Agrawal believes we’ll see a significant increase in AI agency within the next year as these tools become more capable.
Agrawal’s pivot from social media drama to practical AI research seems smart. While Twitter burned, he was apparently building something that could revolutionize how AI agents interact with information. Meanwhile, he’s still locked in a legal battle with Elon Musk over $50 million in unpaid severance from his Twitter departure. Not a bad comeback story.
