Firecrawl MCP, Firecrawl's official web scraping server

• Reads one web page and returns its content

  Scrapes a single known URL and returns the content, preferably as structured JSON or else markdown. Use it when you know the exact page you want.

  Try: “Scrape https://example.com/pricing and give me the plan names and prices”

  Network · firecrawl_scrape

• Searches the web and can return page content

  Runs a web search and returns results, with options to restrict by category or source. Use it when you do not know which page has the answer.

  Try: “Search the web for the latest Next.js release notes”

  Network · firecrawl_search

• Lists the URLs found on a website

  Discovers URLs on a site so you can see what pages exist before reading any of them. Use it to find the right page on a large site.

  Try: “Map docs.example.com and list the pages about authentication”

  Network · firecrawl_map

• Reads many pages across a site or section

  Crawls a site or section and returns the extracted pages once the crawl finishes. Always set limits so it does not run wide. Use it to pull a whole docs section.

  Try: “Crawl the first 20 pages of docs.example.com/api”

  Network · firecrawl_crawl

• Clicks, types and navigates on a live web page

  Opens a URL, or continues on a page already scraped, and lets the assistant click, type and navigate to reach content that needs interaction. Use it for dynamic pages.

  Try: “Open the site, click the Pricing tab and read the annual prices”

  Runs code · firecrawl_interact

• Turns an uploaded file into markdown or structured data

  Parses files such as PDFs, Word documents, spreadsheets and HTML into markdown or JSON. Pass one file per call. Local file reads need a self-hosted setup. Use it to extract text from documents.

  Try: “Parse this PDF report and give me the key tables as JSON”

  Files · firecrawl_parse

• Researches a question across unknown sites and returns structured data

  Starts an agent that searches several sites on its own and returns structured data, such as a list or a set of fields about an entity. Use it when you do not know which pages hold the answer.

  Try: “Find the founders and funding of 10 AI observability startups”

  Network · firecrawl_agent

• Searches issues, pull requests, READMEs and docs for coding answers

  Searches an index built for coding agents that covers GitHub issues, merged pull requests, READMEs and code documentation. Use it for library behavior, API contracts, error messages and known bugs.

  Try: “Why does this library throw a ECONNRESET error on retry?”

  Network · firecrawl_developer_search

• Searches US government law and regulation sources

  Searches statutes, regulations, codes, court opinions and other US federal, state and local government publications. Use it when you need the governing text or an official source.

  Try: “Find the federal regulation on data breach notification timelines”

  Network · firecrawl_gov_search

• Searches scientific paper abstracts

  Searches paper metadata and abstracts with a natural-language query, with optional author, category and date filters. Use it to start a literature search.

  Try: “Find recent papers on GLP-1 agonists and kidney outcomes”

  Network · firecrawl_research_search_papers

• Gets the details of one paper by its ID

  Retrieves the canonical metadata for a single paper from an arXiv, PMC, PMID or DOI identifier. Use it to confirm what a specific paper is.

  Try: “Show details for the paper with DOI 10.1000/xyz123”

  Network · firecrawl_research_inspect_paper

• Finds papers connected to ones you already have

  Expands from one or more anchor papers through the citation graph to find related work. Use it to widen a literature review.

  Try: “Find papers related to arXiv 1706.03762”

  Network · firecrawl_research_related_papers

• Reads passages from a paper's full text

  Reads full-text passages from a specific paper. Use it when the abstract is not enough and you need the detail.

  Try: “Read the methods section of that paper”

  Network · firecrawl_research_read_paper

• Sets up a recurring check on a web page

  Creates a monitor that rescans one or more pages on a schedule and compares each result with the last snapshot. A plain-English goal tells it which changes matter. Use it to watch pages for updates.

  Try: “Monitor example.com/pricing and tell me if the prices change”

  Write · firecrawl_monitor_create

• Lists your page monitors

  Lists the monitors on your account. Use it to see what you are already watching.

  Try: “List my web page monitors”

  Network · firecrawl_monitor_list

• Shows the details of one monitor

  Returns the settings and state of a single monitor. Use it to check how a monitor is configured.

  Try: “Show the details of my pricing monitor”

  Network · firecrawl_monitor_get

• Changes a monitor's goal or alert settings

  Updates a monitor, including its goal, whether change judging is on, and its webhook or notification settings. Use it to tune what you get alerted about.

  Try: “Update my pricing monitor to only alert on price changes”

  Write · firecrawl_monitor_update

• Runs a monitor check right now

  Triggers an immediate check instead of waiting for the next scheduled run. Use it to test a monitor or get fresh results.

  Try: “Run my pricing monitor now”

  Runs code · firecrawl_monitor_run

• Removes a monitor permanently

  Deletes a monitor. The docs mark it as destructive, so only use it when you mean to remove it.

  Try: “Delete my old pricing monitor”

  Delete · firecrawl_monitor_delete

• Lists the past checks of a monitor

  Lists the checks a monitor has run, optionally filtered by status. Use it to see the history of runs.

  Try: “Show the last checks for my pricing monitor”

  Network · firecrawl_monitor_checks

• Shows what changed in one monitor check

  Returns page-level results for a single check, including the diff, the snapshot and whether the change was judged meaningful. Use it to see exactly what changed.

  Try: “Show what changed in the latest pricing check”

  Network · firecrawl_monitor_check

• Shows your remaining Firecrawl credits

  Returns your remaining and plan credits for the billing period, or historical usage by month and by API key. It needs an authenticated account. Use it to check spend.

  Try: “How many Firecrawl credits do I have left?”

  Network · firecrawl_credit_usage

• Used by the assistant behind the scenes

  Checks how far a crawl job has got

  Looks up the status of a crawl that was started earlier. Use it when a crawl is still running and you need its progress or results.

  Try: “Check the status of my crawl job”

  Network · firecrawl_check_crawl_status

• Ends a browser session started for a page

  Stops an interact session for a scraped page once you are done. Use it to clean up after browsing a page.

  Try: “Stop the browser session on that page”

  Runs code · firecrawl_interact_stop

• Checks the progress of a research agent job

  Looks up the status and results of an agent job started earlier. Use it to fetch the answer once the agent finishes.

  Try: “Check whether my research agent has finished”

  Network · firecrawl_agent_status

• Browses the Alexandria catalogue of data providers

  Browses the catalogue of data providers step by step, from categories down to the exact tool contract. It runs nothing and costs no credits. It needs an API key on a team with Alexandria access.

  Try: “What data providers does the Alexandria catalogue have for podcasts?”

  Network · firecrawl_find_tools

• Sends a quality rating on a search result back to Firecrawl

  Sends structured feedback on an earlier search. The first feedback per search refunds one credit. It can be disabled with an environment variable. Use it after using a search.

  Try: “Send feedback that the last search missed the official docs”

  Write · firecrawl_search_feedback

• Sends feedback on a finished scrape, parse, map or search job

  Sends structured feedback for a completed endpoint job. For search quality specifically, the search feedback tool is preferred.

  Try: “Send feedback on that scrape job”

  Write · firecrawl_feedback

Watch out for

• Local file read through parse in self-hosted setups

  When the local server points at a self-hosted Firecrawl, the parse tool reads any file path the model passes, including files outside your project. Two reports are still open, so keep the model away from sensitive folders. github.com

• Search is marked read-only but can click and type

  The search tool tells clients it is read-only, yet its scrape options accept browser actions such as clicking, typing and running JavaScript. A client may skip asking you first, so review these calls. github.com

• Agent runs and crawls can spend many credits

  An agent run may spend up to 2,500 credits by default, and crawls bill per page. Set a credit limit and crawl limits, and use a key with a pay-as-you-go cap. github.com

• An exposed local HTTP server shares your API key

  If you run HTTP mode on 0.0.0.0, anyone who can reach the port uses your Firecrawl key. Keep it on localhost. github.com

• Pages are fetched by Firecrawl's cloud

  With the hosted address or a cloud API key, every URL and query goes to Firecrawl's servers. Self-host Firecrawl if pages or queries must stay in-house. docs.firecrawl.dev

• The bundled zod library has a high-severity advisory

  The Firecrawl SDK the server pins pulls in zod 3, which npm audit flags for a memory-exhaustion advisory with no 3.x fix. It is waiting on an SDK release. github.com

Security advisories

In their words