As of 16 November 2026, Firecrawl MCP offers free, limited web search, page reading and document parsing without an account, useful for workers connecting an AI assistant to outside information. Signing in is not the same as paying: Firecrawl’s free account includes 1,000 credits a month. A paid plan buys more capacity, while some data-provider tools need separate eligibility checks.
MCP, short for Model Context Protocol, lets an assistant connect to external tools. Firecrawl announced MCP v3 on 12 November. For someone choosing a connection now, the important distinction is between accountless access, a free account and a paid plan.
What can you do without a Firecrawl account?
The hosted keyless connection needs neither an account nor an API key. It exposes three tools: Search finds web results, Scrape retrieves page content, and Parse extracts content from documents. It does not expose the wider authenticated catalogue.
Keyless use has two daily caps per IP address—one on requests and another on credits consumed. Reaching either returns a limit error, HTTP 429. The documentation does not state numerical ceilings or a reset timezone for this hosted keyless allowance. That makes it a starting point for occasional tasks, not a published capacity guarantee for recurring work.
Because the limit is per IP, colleagues sharing an office’s public internet address may share the allowance. That is an implication of the rule, not a separate allowance per employee.
Document access also has a practical boundary: the hosted server cannot read a file directly from your computer. Its Parse workflow requires an upload handoff before processing the uploaded file.
What does signing in unlock on the free plan?
Browser sign-in and API-key authentication reach the same tool catalogue, subject to the team’s plan and policy. Neither authentication method is itself a paid tier.
Firecrawl’s Free plan includes core tools such as Map, Crawl, Interact and Monitor: finding a site’s URLs, collecting multiple pages, interacting with a browser and checking for changes. If Crawl is missing from an accountless connection, the first step is to connect an account and inspect the tool list—not assume a subscription is necessary.
The free account needs no payment card. Its monthly credits reset each cycle, and unused credits do not carry over. For ordinary credit-consuming requests that exhaust the balance, Free returns HTTP 402 until the reset; it has no automatic pay-as-you-go top-ups. This balance rule should not be treated as an explanation of the separate daily MCP offer below.
When does a paid plan help?
The published USD prices list Hobby at US$19 with monthly billing, including 5,000 credits each month. The displayed US$16 monthly equivalent requires annual billing; it is not a US$16 month-to-month subscription.
Payment also increases simultaneous work and request rates. The plan limits list two concurrent browsers on Free and five on Hobby. For each of Search, Scrape and Map, the published API rate rises from 10 to 100 requests per minute. These rate counters are shared across a team’s keys. A workload can fit the monthly credit balance but still need slower scheduling to stay within its rate limits.
Paid self-serve plans can enable automatic top-ups in US$5 increments with a monthly spending limit. That is an additional spending choice, not part of the free account.
Does daily free MCP usage replace monthly credits?
Firecrawl’s November 12 announcement says every plan includes free MCP usage replenished daily. It does not quantify that allowance, specify every covered tool or explain how it interacts with monthly credits and provider charges.
Keep three things separate: the accountless per-IP daily caps, the account’s monthly credit balance, and the advertised daily MCP allowance. The public information does not establish how to combine them into a dependable free-usage budget. Treat the daily MCP offer as unquantified when planning recurring work.
Would 1,000 credits cover a small team’s research?
Consider an illustrative workload: ten research requests per working day for twenty days, or 200 requests. Assume each gets one to ten ordinary web results, then retrieves exactly three HTML pages. This is a credit calculation, not a measured bill.
The standard Search rate is two credits per started group of ten results; retrieving the pages adds scrape charges. A basic scrape costs one credit per page, with four more for JSON extraction—turning page content into structured fields.
Plain page text: 200 × (2 search credits + 3 × 1 scrape credit) = 1,000 credits.
JSON extraction on all three pages: 200 × (2 + 3 × (1 + 4)) = 3,400 credits.
The first workload exactly consumes the listed free monthly balance, leaving no room for other credit-billed work. The second exceeds Free but fits Hobby’s 5,000-credit allowance. Neither figure is a separately purchasable job price or a guarantee of MCP billing.
Assumptions exclude retries, PDFs, provider calls, other surcharges and other account activity; assistant or model fees and taxes are outside the calculation. No unquantified daily MCP allowance is deducted. Choosing plain text when structured fields are unnecessary materially changes this budget.
Why Alexandria needs a separate access check
Alexandria is Firecrawl’s catalogue of data tools. It requires an authenticated MCP session; keyless sessions do not get it. Signing in is necessary, but does not establish that every provider is included on Free.
According to Alexandria’s pricing and access rules, browsing the tool catalogue is free; execution follows each tool’s listed price. A discovery request that also searches the web can still incur web-search charges. Government and Research Index tools are documented free exceptions with their own call limits. Some providers also require an organization administrator to review and accept terms before use.
For a provider-backed app, check plan eligibility, execution price and permitted reuse separately. RohitAI’s guides to Firecrawl company-data rights and costs and Benzinga display permissions for AI apps cover those different provider decisions without assuming one provider’s rules apply to another.
Choose the connection before buying capacity
Start keyless if Search, Scrape and Parse cover an occasional task. Connect a free account when you need additional tools or the published monthly allowance. Consider paying when the expected workload exceeds that allowance, needs more simultaneous processing, or requires a provider capability with paid eligibility.
Use the official keyless setup or browser sign-in guide for your client, then check its connected tool list. Browser sign-in requires a human to approve the Firecrawl team. Confirm the intended tool is available and check account usage before scheduling recurring work.
