Celts, which develops search infrastructure for AI agents, has raised $12.5 million in seed funding. SiliconANGLE reported on June 25 that Celts plans to use the funds to expand search infrastructure that helps AI agents find information on the web.

The funding round was led by Speedinvest and B Capital. Italian Founders Fund, Future Back Ventures, Future Present, Arc Investors, Bento Ventures, Mango Capital, 2100 Ventures and United Ventures also participated. Angel investors from Google, Hugging Face and Ramp Network also joined the round.

Celts is focusing on building underlying search infrastructure rather than an AI answer engine like Perplexity. The company says existing search methods are designed around people and do not fit AI agents that process long and complex queries in parallel.

Antonio Maglia (안토니오 말리아), CEO of Celts, said, "Much of the information AI agents need sits in areas existing search engines do not find well. Page-level information such as sentences in the body text, tables, images and snippets is a representative example."

Celts built its entire search stack in-house, from crawlers and indexes to search models and ranking systems.

Celts built a news index first and released its platform 8 months after starting development. In its in-house 'Dynamic News Search' benchmark, it reported 89 percent accuracy and response times of less than 250 milliseconds. The figures are not from an independent evaluation, but Celts said it crawls hundreds of millions of pages a day and typically returns results within 200 milliseconds.