Sign in Book a demo
Back to Tools
Developer Tools API Key BYOCPro
Build Your Agent View Firecrawl Docs
firecrawl logo

Firecrawl MCP Tools for AI Agents

Let an agent search the web, read known pages, crawl sites, and extract structured records from plain-language instructions.

Available Tools

12 tools available in this integration

Firecrawl.CancelCrawl

Stop a crawl that is still running.

Firecrawl.CrawlWebsite

Read many pages of one website and return each page's content.

Firecrawl.ExtractData

Collect structured records from the web from a plain-language description.

Firecrawl.GetCrawlData

Read the pages a crawl has collected so far.

Firecrawl.GetCrawlStatus

Check how far a crawl has progressed, without fetching its pages.

Firecrawl.GetExtractStatus

Collect the results of an extraction that is already running.

Firecrawl.MapWebsite

List the URLs on a website, without reading the pages.

Firecrawl.ScrapeUrl

Read one web page whose URL is already known and return its content.

Firecrawl.Search

Search the web and optionally read each result page in the same call.

Firecrawl.SearchDeveloperDocs

Search developer documentation and repositories, quoting the text that matched.

Firecrawl.SearchGithubIssues

Search GitHub issues across indexed repositories.

Firecrawl.SearchResearchPapers

Search published scientific papers and return their abstracts.

Why agent-optimized tools matter

Built for intent, not API schemas

Ask the agent to "find every page about pricing and summarize the current plans." Arcade helps it choose the right Firecrawl actions without guessing crawl IDs, page URLs, or extraction settings.

Up to 100x fewer tokens

Firecrawl search, mapping, and status checks return lightweight results, while full page content and extracted records are fetched only when needed. The agent avoids loading entire websites into the model just to choose its next step.

Complete the workflow end to end

Search the web, map a site, crawl its pages, extract structured records, and return cited findings through one toolkit designed around the complete web research task.

Make actions predictable and controlled

Arcade tools return structured results and useful errors across users and runs. Teams can apply scope restrictions, redaction, and approval policies so agents follow the same controls every time.

Securing Firecrawl

The more capabilities agents get, the more useful they are. However, more capabilities bring more risks of undesired effects such as data loss and exfiltration.

This is why Arcade provides you with contextual rules that dictate how and when a tool is allowed to run. Arcade’s contextual access lets you govern tool visibility and behavior at every stage of execution and control who can see a tool, what inputs are allowed, and what comes back.

Learn about Contextual Access Policies

For Firecrawl, you can implement some of these policies:

Access when tools are listed for a user
  • Hide Firecrawl.CancelCrawl from agents outside the roles your IDP marks as crawl operators, so regular agents cannot stop another team's collection job.
  • Restrict Firecrawl.CrawlWebsite to agents operating under your approved research role, so unreviewed automation cannot collect an entire site.
  • Deny Firecrawl.ExtractData to agents that lack approval to gather structured records from external websites, so broad data collection cannot run under a general-purpose role.
Pre-execution before each tool call
  • Require human approval before Firecrawl.CancelCrawl runs, so an agent cannot discard an in-progress crawl without an operator's sign-off.
  • Block Firecrawl.CrawlWebsite when its target is outside your domain allowlist, so an agent cannot crawl customer, partner, or internal sites by mistake.
  • Refuse Firecrawl.ExtractData requests that target domains outside your approved sources or ask for personal data, so an agent cannot turn arbitrary websites into people records.
Post-execution after the tool returns
  • Redact Firecrawl.GetCrawlData results for email addresses, phone numbers, access tokens, and other credentials before the page content reaches the model.
  • Strip API keys, session tokens, passwords, and personal contact details from Firecrawl.ScrapeUrl results, so a readable page cannot expose secrets to the model.
  • Deny Firecrawl.Search results when a returned page contains credentials or data under legal hold, so restricted web content never reaches the model.

How to use Firecrawl with your agent

Connect Firecrawl to an MCP client and give your agent web search, page reading, crawling, and extraction in a few steps. Pick your client below to follow the setup.

Your questions, answered

A Firecrawl MCP server exposes web-reading and web-research capabilities as structured tools an AI agent can call through the Model Context Protocol. The agent can search the web, read pages, crawl sites, extract structured records, and inspect developer docs, GitHub issues, and research papers.

Firecrawl's own MCP server is a Firecrawl endpoint for calling its web data capabilities. Arcade adds brokered API-key authentication, approval workflows, response redaction, tool controls, and access through the MCP clients and agent frameworks your team already uses.

Arcade agents can search the web, read known URLs, map websites, crawl multiple pages, check crawl progress, retrieve crawl results, extract structured records, cancel running crawls, and search developer docs, GitHub issues, and research papers.

Arcade uses the shared Firecrawl API key or secret configured for the integration. The key determines which Firecrawl capabilities and account quotas the agent can use; this integration does not use per-user OAuth scopes.

No. Firecrawl uses one shared API key or secret configured once for the integration, rather than having each user connect an individual account. Arcade brokers that credential, and the model never receives it.

Yes. Contextual Access policies can require human approval before consequential Firecrawl actions, such as starting a crawl, extracting records, or canceling a running crawl. The call proceeds only after the policy's conditions have been satisfied.

Arcade proxies Firecrawl API calls and does not persist your content as part of normal tool execution. The shared credential is brokered and stored encrypted at rest, and the model never receives it.

Any framework or client that can connect to Arcade's MCP gateway or tool APIs works, including LangChain, OpenAI Agents, CrewAI, Claude Desktop, Claude Code, Cursor, Visual Studio Code, Microsoft Copilot Studio, and GitHub Copilot.

Arcade can continue using the Firecrawl key while it remains valid. If the key is revoked, rotated, or can no longer authenticate, calls stop until the integration is reconfigured with a valid credential.

Firecrawl's rate limits, crawl and extraction limits, and account quotas still apply to calls made through Arcade. Exact limits depend on the Firecrawl account and plan; Arcade does not remove those service constraints.

Ready to actually ship your agent?

Connect your agent to Firecrawl with a shared API key, agent-optimized tools and approval controls for consequential actions.

Build Your Agent Read the Firecrawl Quickstart