01
Overview
Firecrawl launched Alexandria on September 22, 2026 as a knowledge layer for AI agents. Instead of forcing an agent to choose separately between web search, scraping, a specialist API and a private index, Alexandria is designed to give it one way to discover a source, inspect what that source contains and retrieve the relevant information.
The launch combines four broad source types: the live web, official data providers, site-specific or custom connectors, and Firecrawl's own specialized indexes. At launch, Firecrawl highlights Research, Developer and Government indexes. Its live homepage currently reports 113M+ indexed sources, 82 providers, 471 capabilities and 28 categories; these are live catalog figures and can change.
What makes Alexandria different from normal Firecrawl search? Firecrawl Search finds relevant pages on the live web and can return their content. Scrape turns a known URL into clean Markdown or structured data. Alexandria expands the retrieval surface beyond ordinary webpages. An agent can reach structured provider capabilities and specialized indexes as well as the live web, while keeping them inside the Firecrawl workflow.
02
How it works
Firecrawl says its Research Index contains tens of millions of scientific-paper abstracts, its Developer Index spans tens of millions of primary sources including documentation, READMEs, issues and merged pull requests, and its Government Index covers laws, regulations and ordinances. Those descriptions come from Firecrawl's launch materials; the exact coverage will evolve as the indexes are updated.
How do you access Alexandria? Firecrawl says developers can connect an AI agent through Firecrawl's MCP or build with the API. Its CLI also exposes Alexandria provider/capability addresses. Current CLI documentation shows that a provider capability can be supplied where a URL would normally appear, while the explicit --alexandria form remains supported. Tool addresses are validated as Alexandria capabilities and do not silently fall back to URL scraping.
A practical example is technical research. An agent could use the Developer Index for primary-source documentation, READMEs, issues and merged pull requests, then use live-web search for current announcements. Scientific research can similarly combine the Research Index with live sources. This matters because a retrieval system is often limited less by the language model than by whether it can find authoritative, current material in a predictable form.
03
Implementation
Firecrawl reports that agents using Alexandria scored 21% higher on answer quality than agents using built-in web tools across 845 tasks, using the same model and prompts with blind AI judging. Treat this as a vendor-run benchmark, not an independent evaluation. It is evidence about Firecrawl's own test setup, not proof that Alexandria will improve every agent or workload by 21%.
Pricing is less settled. Firecrawl's general service uses credits and publishes pricing for Search, Scrape, Crawl, Interact and Agent, but the Alexandria launch post does not establish a simple universal standalone Alexandria price. The homepage says the Research Index paper endpoints are free across categories, while other provider capabilities can have their own pricing. Check the live catalog and billing documentation before estimating production cost.
Alexandria also has a provider-side ambition. Firecrawl says it already pays official data providers through individual agreements, including Wikimedia Enterprise, and plans a self-service system that could let individuals, creators and organizations earn when agents use their knowledge. That self-service system is described as coming soon, so it should not be presented as generally available today.
04
Limits and verification
Who is Alexandria for? The strongest fit is teams building research agents, coding agents, RAG systems, competitive-intelligence workflows, lead research, compliance research and other products where source quality matters. It is less relevant when an application already has a known URL and only needs a single scrape; ordinary Firecrawl Scrape remains the simpler primitive for that job.
The early takeaway is that Alexandria is not a new foundation model. It is retrieval infrastructure: a unified source and capability layer intended to give AI agents broader access to authoritative and specialized information. Its search opportunity is therefore likely to center on how to use it, API and MCP access, providers, indexes, pricing, comparisons with web-search APIs, and whether its retrieval quality holds up outside Firecrawl's own benchmark.
FAQ: Is Alexandria available now? Firecrawl says Alexandria starts now. Is it the same as Firecrawl Search? No; Search focuses on the live web, while Alexandria adds official providers, connectors and specialized indexes. Can agents use it through MCP? Firecrawl explicitly says yes. Is the 21% quality improvement independently verified? No; it is a Firecrawl-run evaluation. Is provider monetization live for everyone? No; Firecrawl says the self-service provider system is planned to open soon.
Sources
Primary and supporting sources
Use the linked documentation to check current pricing, access and product behaviour. The dates above describe this article; they do not guarantee that every vendor claim remains current.