Parallel Web Systems — $2B in five months on agent search APIs
Parag Agrawal isn't running Twitter anymore. He's running a startup that just tripled its valuation in five months.
Parallel Web Systems closed a $100M Series B Tuesday at a $2B valuation, led by Sequoia, with Kleiner Perkins, Index Ventures, Khosla Ventures, First Round Capital, Spark Capital, and Terrain Capital participating. Five months ago they raised $100M Series A at $740M led by Kleiner and Index. Total raised: $230M.
What they sell: web search and research APIs designed specifically for AI agents. Customers named in the announcement include Clay, Harvey, Notion, Opendoor, plus undisclosed banks and hedge funds. Over 100,000 developers on the platform.
The thesis worth pulling out: while everyone fights over the agent client layer, the agent tool layer is quietly compounding. Web search is the highest-frequency capability an agent uses, and SerpAPI, Bing, and Google APIs all return human-readable SERPs, not agent-consumable structured data. Parallel is the cleanest pure-play on search rebuilt for agents instead of for humans.
Sequoia's pace on agent infrastructure has accelerated this month. Cognition's parent on the Devin Terminal launch, now Parallel Series B at $2B. The bet is explicit: in this wave, the high-margin layer above the foundation models is the tool stack agents call. Parag's Twitter-era data infrastructure DNA maps directly here. The web stops being something humans browse and becomes something programs consume.
Link: https://parallel.ai
← Back to all articles
Parallel Web Systems closed a $100M Series B Tuesday at a $2B valuation, led by Sequoia, with Kleiner Perkins, Index Ventures, Khosla Ventures, First Round Capital, Spark Capital, and Terrain Capital participating. Five months ago they raised $100M Series A at $740M led by Kleiner and Index. Total raised: $230M.
What they sell: web search and research APIs designed specifically for AI agents. Customers named in the announcement include Clay, Harvey, Notion, Opendoor, plus undisclosed banks and hedge funds. Over 100,000 developers on the platform.
The thesis worth pulling out: while everyone fights over the agent client layer, the agent tool layer is quietly compounding. Web search is the highest-frequency capability an agent uses, and SerpAPI, Bing, and Google APIs all return human-readable SERPs, not agent-consumable structured data. Parallel is the cleanest pure-play on search rebuilt for agents instead of for humans.
Sequoia's pace on agent infrastructure has accelerated this month. Cognition's parent on the Devin Terminal launch, now Parallel Series B at $2B. The bet is explicit: in this wave, the high-margin layer above the foundation models is the tool stack agents call. Parag's Twitter-era data infrastructure DNA maps directly here. The web stops being something humans browse and becomes something programs consume.
Link: https://parallel.ai
Comments