Connect your AI Agents to Firecrawl in minutes

What is a Firecrawl MCP server?

It’s an MCP server that lets you access a specific instance of Firecrawl via tools. Your agents can call these tools to scrape specific webpages, search the web with a query, return the status of a crawl job, and more.

Firecrawl offers their own MCP server. But you can access additional ones through 3rd-party platforms, like Merge Agent Handler.

How can I use the Firecrawl MCP server?

Here are a few common use cases:

• Web crawling for RAG: Crawl a site (or a set of pages), extract clean text/markdown, and feed it into your retrieval pipeline
  

• Competitor/market research: Crawl competitor docs, pricing pages, changelogs, and blog posts, then summarize what changed and what matters
  

• Support deflection / help center assistant: Index a help center and let an agent answer customer questions with citations to the exacted pages
  

• Content monitoring: Re-crawl specific pages on a schedule and send alert when key sections change (pricing, policy, release notes, etc.)

What tools do Firecrawl MCP servers often expose?

Here are just a few impactful tools:

• <code class="blog_inline-code">Map_site</code>: Crawls a website to discover and return a list of all accessible URLs

• <code class="blog_inline-code">start_crawl</code>: Initiates an asynchronous crawl job to scrape multiple pages across a site

• <code class="blog_inline-code">get_crawl_status</code>: Retrieves the progress and results of an ongoing or completed crawl job

• <code class="blog_inline-code">start_extract</code>: Starts an AI-powered extraction job to pull structured data from one or more URLs

• <code class="blog_inline-code">get_extract_status</code>: Retrieves the status and results of an AI extraction job, including structured output

• <code class="blog_inline-code">start_batch_scrape</code>: Launches an asynchronous batch scraping job across multiple URLs for improved efficiency

What makes Merge Agent Handler’s Firecrawl MCP server better than alternative Firecrawl MCP servers?

Merge Agent Handler provides several platform-level advantages:

• End-to-end authentication and credential management: Merge handles authentication flows, credential storage, and token lifecycle automatically, reducing integration overhead and ensuring secure access to third-party tools
  

• Enterprise-grade security with DLP enforcement: All tool inputs and outputs are scanned through a security gateway that can block, redact, or flag sensitive data based on configurable rules
  

• Real-time observability and fully-searchable logs: Every tool call and underlying API request is logged, enabling teams to monitor behavior, debug issues, and maintain a complete audit trail with searchable logs
  

• Prebuilt connectors and extensibility via Connector Studio: Developers can use managed connectors or customize and extend tools through Connector Studio with full control over integration behavior
  

• Evaluation Suite for pre-production validation: Teams can test and validate tool outputs against expected results before deploying to production, improving reliability and performance

Can I set custom security rules for Firecrawl tool calls in Merge Agent Handler?

Yes, Merge Agent Handler includes a security gateway with DLP that allows you to define custom rules for tool inputs and outputs.

Here are some examples:

• Prevent sensitive-data exfiltration via URLs: Block Firecrawl calls when the target URL matches private/internal domains (e.g., *.corp, staging environments, admin panels) to prevent crawling non-public content
  

• Stop credential leakage into crawl requests: Block tool calls if inputs include API keys, bearer tokens, passwords, or cookie strings (common when someone pastes a full header dump or authenticated URL)
  

• PII guardrails in extracted page content: Redact detected PII (emails, phone numbers, SSNs, credit card patterns, etc.) from Firecrawl outputs before they reach the model, so downstream steps can't accidentally summarize or store it
  

• Allowlist-only crawling for compliance: Permit Firecrawl tool calls only for an approved list of domains (e.g., your docs site and specific competitor domains), and block everything else to keep agents “on-policy”
  

• Tenant boundary protection (customer-facing agents): Block crawls that attempt to fetch content from a different customer’s domains or known assets to reduce cross-tenant leakage risk.

How can I start using Merge Agent Handler?

You can follow these steps:

1. Create an Agent Handler account and sign in.

2. Get your Agent Handler API key, since your MCP client will use it as a bearer token to authenticate to the Agent Handler MCP endpoint.

3. Create a Tool Pack for the workflow you want to power, because Tool Packs define which connectors and tools your agent is allowed to use.

4. Add connectors and choose tools inside that Tool Pack (only what you want exposed to the agent).

5. Create a Registered User (or decide on a service account pattern). This is the identity boundary Agent Handler uses to scope credentials and audit logs.

6. Have that Registered User authenticate to the third-party tool(s) via Merge Link (or Magic Link), so Merge stores/uses the end-user credentials without you managing tokens directly.

7. Copy the MCP endpoint URL for your Tool Pack + Registered User (format: <code class="blog_inline-code">.../tool-packs/{TOOL_PACK_ID}/registered-users/{REGISTERED_USER_ID}/mcp</code>).

8. Add that MCP server to your MCP client (Cursor, Claude Desktop, VS Code, Windsurf, etc.).

9. Configure your MCP client to send <code class="blog_inline-code">Authorization: Bearer <AGENT_HANDLER_API_KEY></code> on requests to that MCP URL.

10. Test and validate. Start with <code class="blog_inline-code">list_tools</code> and a simple read action, then verify behavior in Logs/Playground, and tighten Tool Pack scope and security rules before production.