The Pulse
Firecrawl Raises $75M to Build Alexandria for AI Agents
Firecrawl is using its $75 million Series B to expand Alexandria, a unified source layer for AI agents that combines the live web with indexes, data providers and custom connectors.
AI.info Team ·
Firecrawl wants agents to use more than the open web
Firecrawl announced a $75 million Series B on September 22, 2026, as it launched Alexandria, a data library designed to give AI agents access to sources beyond ordinary web search. Smash Capital led the round, with participation from Altos Ventures, Nexus Venture Partners, Y Combinator, Freestyle and Offline Ventures. Firecrawl’s announcement says the company will use the funding to expand Alexandria, improve search and pay more of the people and organizations whose knowledge agents retrieve.
The move puts Firecrawl in a contest over who controls the source layer for agent software. Developers can already connect agents to websites through Firecrawl’s crawling, rendering and parsing tools. Alexandria adds official data providers, custom connectors and Firecrawl’s own indexes to that live web access, giving agents a common way to discover a source, inspect what it contains and retrieve information.
From website crawling to source selection
Firecrawl says more than 1.5 million users now build with its products. The company began with a narrower problem: extracting clean, usable information from websites for AI applications. Its founders previously built Mendable, a documentation chat product, and say they found web data collection to be the most difficult part of that product’s stack.
Alexandria addresses the next problem created by more capable agents. An agent researching a company may need its website, financial data, regulatory filings and specialized datasets. Those sources generally expose different APIs, pricing models and formats, forcing developers to maintain separate integrations. Firecrawl’s pitch is that Alexandria can make the source choice part of the agent’s normal retrieval process instead of a separate engineering project.
“We want to make that knowledge available and give the people who contribute it a reason to keep doing so,” — Caleb Peffer, CEO of Firecrawl
The company says that reason will include payment when AI agents use contributed knowledge.
Alexandria starts with indexes and providers
The initial system combines the live web with Firecrawl’s Research, Developer and Government indexes. The Research Index contains tens of millions of scientific paper abstracts. The Developer Index covers tens of millions of primary sources, including documentation, README files, issues and merged pull requests. The Government Index covers laws, regulations and ordinances.
Firecrawl also says agents can query official providers and use specialized tools to collect entire datasets. The company already has agreements with several providers, including Wikimedia Enterprise. According to Firecrawl, millions of requests for Wikipedia data pass through its platform each month, and the company pays for direct access.
That provider model is central to Alexandria’s funding plan. Firecrawl says it is developing a self-service system for individuals, content creators and organizations to earn money when agents use their work. The company plans to open that system soon and says it could eventually include expertise that has never been published online.
A 21% quality gain in Firecrawl’s test
Firecrawl reports that agents using Alexandria scored 21% higher on answer quality than agents using built-in web tools across 845 tasks. The company says it used the same model and prompts in both conditions and relied on blind AI judging. Firecrawl does not describe the evaluation as an independent benchmark, so the result is best read as an internal comparison of its retrieval system rather than a general measure of agent performance.
The company’s broader product challenge is coverage. A search system can only improve an answer when it reaches the information needed to produce it. Alexandria’s combination of indexes, providers and connectors is intended to reduce the number of cases in which an agent stops at a public webpage while the relevant evidence sits in a filing, a dataset or a specialized source.
The business behind the library
Firecrawl says it will spend the new capital on deeper indexes, retrieval systems, live-web access and additional first-party sources. Alexandria is available through Firecrawl’s API, MCP connection and command-line tools, allowing developers to add it to existing agent workflows rather than adopt a separate research application.
The financing also gives Firecrawl room to build a marketplace around agent-mediated knowledge access. Its immediate proposition is technical: one interface for web pages, indexes, providers and connectors. Its longer-term proposition is economic: data owners and individual experts should receive compensation when agents retrieve what they know.
Alexandria launches on September 22 alongside the funding announcement. Firecrawl’s next test is whether its unified access layer can attract enough high-value providers—and enough agent traffic—to make that payment system useful beyond the company’s existing agreements.