Firecrawl Review (2026): Alexandria, the Free Government Index, and Whether 4.6 Holds

Should your agent buy its data through Firecrawl, or keep paying each provider directly? Since October 1, Alexandria has put Apollo, FullEnrich, and Data Legion enrichment on the same key and credit balance as scraping, and on October 22 Firecrawl opened a free index of 30 million US government pages (changelog). Plan prices did not move. Firecrawl also published a benchmark ranking the same enrichment providers it now resells.

Our pick, by buyer. A team with no enrichment vendor should try Apollo through Alexandria before buying an Apollo seat. A team already paying Apollo for sequencing gains almost nothing by switching. Compliance, tax, and permitting agents should wire in the Government and Legal Index today, because searching it is free. Everyone else picks a tier exactly as before. The rating holds at 4.6.

One correction: our September 4 version listed Standard, Growth, and Scale at 50, 100, and 150 concurrent browsers, and the live pricing page (effective September 4, 2026, fetched October 23) says 25, 50, and 100.

Alexandria in brief

Alexandria is a catalogue of data tools an agent reaches through its existing Firecrawl key. An agent searches with alexandria as a source, gets back ranked tools with their inputs and prices, then runs the chosen tool through the scrape endpoint. Discovery is free and execution "uses the credits listed on the tool," per the Alexandria docs. Firecrawl launched it on September 22 alongside a $75M Series B led by Smash Capital, per its launch entry.

Firecrawl's headline quality number comes from that launch entry: agents using Alexandria scored 21% higher on answer quality than agents using built-in web tools, across 845 tasks with blind AI judging (launch entry). That is Firecrawl's own evaluation, and the September 28 review explains why its task design favors Alexandria.

How big is the catalogue? Firecrawl's own pages disagree. The Google Cloud page says 100+ official data providers. The October 8 Universal Scrape post says more than 200. The Alexandria catalogue showed 331 entries under its All filter when we loaded it on October 23, a count that also includes Firecrawl's own indexes. Any of those counts is a lower bound that grows weekly. Whether the five or six providers your agent needs are in it matters more than the total.

Here is the discovery and execution loop from the docs, using the CLI:

# From the Alexandria docs (docs.firecrawl.dev/features/alexandria)
npm install -g firecrawl-cli
firecrawl login

# Find tools alongside web results (discovery is free; web results bill normally)
firecrawl search "podcast conversations about AI agents" --sources web,alexandria --domain-tools --limit 2

# Inspect inputs and pricing before running anything
firecrawl find-tools --providers particle --level tools --expand options,response

# Run the tool; this bills at the tool's listed price
firecrawl scrape particle/podcasts/episodes/search --options '{"keyword_search":"AI agents","limit":2}'

What shipped October 1 to 22

Nine changelog entries landed between October 1 and 22. The table pulls each one from its own changelog page, with the part that changes a purchase decision.

Also shipped: fantasy sports data for ChatGPT and Claude on October 6 (changelog), a consumer use case with nothing in it for a developer buyer.

Three of them change who should buy Firecrawl: enrichment (with the benchmark behind it), the Government and Legal Index, and Google Cloud Marketplace. Universal Scrape is plumbing that makes the first one cheaper to adopt, since provider calls go through the endpoint you already call and bill "in Firecrawl credits at each tool's listed price," per the Universal Scrape post.

One detail in the October 20 release is easy to misread. The "1,000 free credits a month" in the People Data entry is the same 1,000 a month the Free plan lists on the pricing page. The entry does not say it is a separate pool for people data, so plan as if enrichment calls and scrapes draw from the same 1,000.

Enrichment through Alexandria, or Apollo direct

The October 21 benchmark is the useful artifact here, because it answers "which enrichment provider fills in a company record correctly" with an answer key instead of a vendor claim. Firecrawl gave six providers only a company's website, one lookup each, for 200 US companies (150 public, 50 private), and scored seven fields against values from SEC filings, federal records, and company websites, frozen before any provider was called (benchmark page, first run Oct 2, PDL added to Alexandria Oct 7). The headline score covers four code-graded fields: headquarters, employee count, founding year, and LinkedIn page.

Read the conflict first. Firecrawl ran the benchmark, and it earns credits on four of the six providers, per the benchmark's own FAQ. What offsets that: the harness and a 102-company public half of the answer key are open source, corrections made after the provider calls are listed in the repository, and any provider can submit for a rerun (benchmark page). The scores can be rechecked against the public half, which makes this a better buying input than most vendor charts.

Three findings change a purchase. Employee count is the hard field: Apollo.io scored 77.2%, and the next provider, Ocean.io, scored 56.5% (benchmark page). For revenue, Apollo.io was within 5% of the filed figure on 82.8% of public companies, while People Data Labs and Ocean.io mostly return ranges and land at 4.8% and 2.1% (benchmark page). Industry flips the order: Ocean.io leads at 69.2% and Apollo.io ranks last at 53.4%, largely because it returns broad labels (benchmark page). Ocean.io is not in Alexandria.

So where does that leave a buyer?

Already paying Apollo directly? Stay there. Apollo's own pricing page sells account-wide plans (Free, Basic, Professional, Custom) that bundle email sequencing with credits for emails and phone numbers, and unused credits do not roll over. If your reps use the sequencer, Alexandria replaces none of it. Routing the same Apollo data through Firecrawl gives you one key and one bill for your agent, and that's it.

No enrichment vendor yet, and the consumer is an agent rather than a rep? Try Alexandria first. Apollo's person search through Alexandria is free and enrichment bills at the listed per-call price, per the October 1 entry, so you can run your actual target list against the provider that topped the benchmark before committing to a seat plan. If the agent outgrows that, Apollo direct is still there.

Segmenting by industry? Neither route is the answer on that field alone. Ocean.io led industry labels in Firecrawl's own run and is only available direct.

Two setup frictions apply to the Alexandria route. Some providers require an org admin to accept their data terms before the first call, and a call returns THIRD_PARTY_DATA_TERMS_REQUIRED until someone does (October 1 entry). Provider usage rules also carry through: FullEnrich's terms prohibit marketing to personal email addresses, per the same entry.

The Marketing Data Pack and GEO work

The October 13 pack bundles Similarweb, Oriane, Firecrawl Trends, Particle, Icons Computer, and Wayback Machine for competitor traffic, rising searches, and content briefs (changelog), which makes it topic research for a team already on Firecrawl. Nothing in the entry tracks whether AI answer engines cite your pages, so for that measurement use the tools in our AI search visibility tools guide.

The Government and Legal Index costs zero credits to search

This is the October release most readers can use today without spending anything. The index covers 30 million pages from federal, state, county, and city government sources: statutes, regulations, court opinions, municipal and zoning codes, permit requirements, fee schedules, tax forms, and procurement notices, per the launch post. Searches use no credits and need no API key to start; a key raises rate limits. Pulling a result's full text with scrape bills at the normal scrape rate (launch post).

Access paths, per the October 22 entry: the ChatGPT and Claude plugins, GET or POST /search/gov, the firecrawl_gov_search MCP tool, the firecrawl gov CLI command, and an Alexandria provider (firecrawl-gov-index) that adds the laws and cases each page cites plus jurisdiction, court, and date filters.

# Keyless Government Index search, from the Oct 22 launch post
curl -s -X POST https://api.firecrawl.dev/v2/search/gov \
  -H 'Content-Type: application/json' \
  -d '{"query": "Austin mobile food vendor permit requirements", "k": 10}'

# Same query from the CLI
firecrawl gov "Austin mobile food vendor permit requirements" --limit 10

Two limits. The index is US-only, and Firecrawl's Government Index docs say a match "is not a guarantee of current governing law," since results can include historical versions, bill text, and third-party reproductions. An agent answering a compliance question still has to check issuer, jurisdiction, and effective date on the source page. For a tax, permitting, or procurement agent, that is a small price for retrieval that costs nothing, and the decision to wire it in takes an afternoon.

The credit math, re-checked October 23

All five plan prices render on the pricing page now, including the Hobby and Standard figures that came back blank behind the JavaScript slider on September 4. We read them on both the Markdown and HTML versions on October 23. The page is stamped effective September 4, 2026, so prices have not moved since before the Series B.

Per-endpoint rates are unchanged on the same pricing page: Scrape, Crawl, and Map at 1 credit per page; Search at 2 credits per 10 results; Interact at 2 credits per browser minute; Monitor at 1 credit per page per check; Agent with 5 free daily runs, then dynamic pricing. Developer Index search costs 2 credits per 10 results, rounded up, per the Developer Index docs, which closes a gap our September version flagged. Research Index paper calls are free, and Government Index searches are free per its docs. Alexandria tool calls are the one cost that lives outside this page entirely.

Map that to a workload. A research agent scraping 200 pages a morning burns 6,000 credits a month, which clears Hobby's 5,000, so Standard is the tier (pricing page). Monitor moves faster than people expect: 50 pages checked hourly is 1,200 credits a day, roughly 36,000 a month, at the published 1-credit-per-check rate. That eats more than a third of Standard's 100,000 before any scraping.

The concurrency correction moves one pick. A team that sized Standard for 50 parallel renders off our September 4 numbers gets 25, so Growth (50 concurrent at $333 billed yearly) is the tier for that load, per the pricing page. Hobby, Standard, and Growth credits do not roll over; Scale carries unused credits for one month. Paid plans can top up in $5 increments (1,000 credits on Hobby, 2,000 on Standard) up to a monthly cap you set, which is the control to set before an agent starts choosing paid Alexandria tools on its own.

Who buys which plan

The enrichment row is the new one, and it carries a condition: give the agent its own API key with a credit limit before it runs unattended, since every Alexandria call bills at a price you only see per tool. For a straight URL-scraping job, Alexandria adds nothing, and our Firecrawl vs Apify vs Browse AI comparison still decides the scraping-layer call. For search APIs, see Firecrawl vs Tavily vs Exa, and for a plan-by-plan breakdown, Firecrawl pricing.

The rating: 4.6 holds

October gave Firecrawl a credible answer to where an agent gets structured data, and a free one for regulatory questions, but Alexandria is still budgeted tool by tool with no row on the pricing page, so the buy decision for most readers is unchanged and the score holds.

For a team building agents that need web pages plus structured data on one key, Firecrawl is the pick in October 2026: start on Firecrawl here, wire in the Government and Legal Index if your agent answers regulatory questions, and give Alexandria its own capped key. Rated 4.6 of 5.