Firecrawl MCP server for your AI

One Firecrawl API key. Sign in at firecrawl.dev, open API Keys and copy a key starting with fc-, or create a new one just for PopMCP. Keys belong to your Firecrawl team and have no access levels to choose, so every tool works with any key. PopMCP checks it by reading your remaining credit balance, and every fetch, crawl, search and watch check after that spends credits from your own plan.

fetch_page, screenshot_page and map_site cost 1 credit. fetch_page_fields costs 5 credits a page, and stealth mode for heavily protected sites also costs 5 a page instead of 1. search_web costs 2 credits per 10 results. crawl_site costs 1 credit per page, with a default of 50 pages and a cap of 500. Each watch check costs 1 credit per page, or 5 with a fields template. The free plan has 1,000 credits a month.

No. Every request is checked before it leaves PopMCP, and custom headers, cookies, Authorization values, saved browser profiles and robots.txt overrides are refused, as are browser actions that type, press keys or run scripts, which is how logins and CAPTCHAs are worked through. Wait, click, scroll and screenshot actions are fine. A URL with a username or password in it is refused too. Public pages only, fetched by your own key.

create_watch registers up to 50 public URLs with Firecrawl, which checks them on its own servers, daily by default. Anything more often than hourly is refused unless you confirm it, because each check spends credits on every page. Give the watch a goal and an AI judge marks each change meaningful or not; templates for pricing pages, homepages and product pages extract fields. get_watch_changes returns the diffs, and pause_watch stops the spending.

One: watched page changed. When a page in one of your Firecrawl watches changes on a scheduled check, Firecrawl sends a signed webhook and PopMCP hands the change to whatever you want to run next, such as a message to your team or a note on a competitor record. Because the checking happens on Firecrawl's side, nothing runs on your machine and nothing is fetched until a check is due.

Firecrawl answers 402 and the tools stop fetching until credits are available again. Free plan credits reset each month and do not roll over; paid plans can turn on pay-as-you-go top-ups with a monthly limit you set. The AI can check the balance with get_credit_usage. Plans also cap request rates: the free plan allows 10 scrapes and 2 crawls a minute, and Firecrawl answers 429 above that.

No. PopMCP hosts it and gives you a Streamable HTTP MCP URL. There is nothing to deploy, no callback URL to register, and no infrastructure to keep running or patch.

Any client that supports a remote Streamable HTTP MCP server: Claude Desktop, Claude Web, Claude Code, ChatGPT custom connectors, Codex and the Codex CLI, OpenCode and Z.ai's ZCode, as well as editors such as Cursor, Windsurf and Zed. Your client signs in to PopMCP with OAuth; the provider credential never leaves the server.

Connect Firecrawl in PopMCP first, then run claude mcp add --transport http popmcp https://app.popmcp.com/mcp in a terminal. Inside Claude Code, run /mcp and sign in to PopMCP in the browser window it opens, ticking Firecrawl on the consent screen. The Firecrawl tools are then available in that session, and no Firecrawl credential is stored on your machine.

Loading 62 tool definitions into every conversation burns context before the work starts. PopMCP loads a curated 12-tool set covering the everyday workflows and keeps the rest searchable, so your AI pulls a specialist tool only when a task needs it. Every tool stays searchable, and on Scale the owner can load the full catalog for a teammate's connection.

A tool is something your AI calls, and there are 62 here. A trigger is something PopMCP watches for you: 1 Firecrawl events that can start work on the PopMCP side. Triggers never enter your AI client's context, and none of them ask you to paste a callback URL into Firecrawl.