MCP.so
登录

Firecrawl MCP Server已精选

@mendableai

基本信息

分类

浏览器自动化

许可证

MIT

传输方式

stdio

作者

mendableai

提交者

Victor Bruno Castro

关于 Firecrawl MCP Server

🔥 官方 Firecrawl MCP 服务器 - 为 Cursor、Claude 及其他任何 LLM 客户端添加强大的网页抓取和搜索能力。

Firecrawl MCP Server 配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "firecrawl-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR_FIRECRAWL_API_KEY"
      }
    }
  }
}

Firecrawl MCP Server 工具

26

Scrape one URL and return its content: markdown by default, or HTML, links, screenshots, branding data, a targeted answer, or JSON matching a supplied schema. Use it when the request identifies a page and needs its content or defined fields: the user gives a URL or names a page or PDF and wants it read or summarized, or wants prices, specs, contact details or a table from it. Use `firecrawl_search` when additional web sources are needed; on an authenticated session, `firecrawl_map` lists a site's URLs and `firecrawl_crawl` collects a set of pages. Firecrawl may serve recently indexed content; set `maxAge: 0` for a live fetch or a smaller `maxAge` to bound staleness. A successful response does not by itself confirm the page is still current. Browser actions can change the live page when interactive actions are enabled. Authenticated responses can include a `metadata.scrapeId` for optional scrape feedback. On an authenticated session, `firecrawl_search` and `firecrawl_find_tools` can discover providers for the same fields across several pages; a matching provider returns typed records in one call. With `alexandria` in place of `url`, the tool runs catalogued provider capabilities instead of fetching a page. Some providers require accepted terms; when a response reports a terms requirement, direct the user to the Firecrawl dashboard URL in that response.

Enumerate URLs indexed under one website through Firecrawl without fetching each page's content. Use this when the request asks for a site's URL inventory, when several relevant pages must be located, or when the desired page URL is unknown. An optional `search` term narrows the URL list, while sitemap, subdomain, query-parameter, and result-limit options control coverage. Returns matching URLs rather than page bodies. Retrieve one page with `firecrawl_scrape`; collect content across multiple pages with `firecrawl_crawl`. Authenticated responses can include an `id` for optional map feedback.

Search web, news, or image sources and return ranked results with query-relevant highlights. Use it for current information, or to list businesses, people, products, jobs or events with websites, prices or contacts. Each web result is a title, URL, and description; use `firecrawl_scrape` on a result URL when the excerpt is not enough. Authenticated search also returns matching Alexandria data providers in data.tools (companies, people, jobs, finance and filings, public records and government spending, real estate, places and restaurants, retail and prices, package registries and developer data, news, research, and more). Prefer a provider over scraping pages when the task needs the same fields across several entities, exact figures or timestamps, provenance, or many records; use web results when they already answer the question. Authenticated searches always add alexandria to the sources, so sources: ["web"] also returns providers; domainTools: false turns off website-matched tools only. On an authenticated session, tool matches describe available capabilities; `firecrawl_find_tools` returns their contracts and `firecrawl_scrape` with an `alexandria` body executes a selected capability. Keyless sessions get no Alexandria matches in data.tools. For a programming question, add `categories: ["developer"]`; its hits return in `data.web` with `category: "developer"`. For a legal or regulatory question, `categories: ["gov"]` returns hits in `data.web` with `category: "gov"` and cannot be combined with other categories; authenticated sessions also have `firecrawl_gov_search` as the dedicated tool. Verify gov results' issuer, jurisdiction, status, and version before citing current law. `categories: ["research"]` restricts web results to research-affiliated websites; the `firecrawl_research_*` tools are a separate surface over paper abstracts and full text (PubMed, bioRxiv, medRxiv, arXiv). Returns source-type result groups and usage metadata. Authenticated responses can include an `id` for optional search feedback.

Browse Alexandria data providers and workflows or read a selected contract. Alexandria covers companies, people, jobs, finance and filings, public records and government spending, real estate, places and restaurants, retail and prices, package registries and developer data, news, research, and more: typed, sourced records through published contracts. Prefer normal firecrawl_search for a data task; it already returns matching providers. Use this tool when the contract you need was not returned in full, to browse a category when search found nothing, or before scraping the same fields from several pages. Discovery is free. Use query for semantic discovery or urls to find providers for a website. With no arguments, browse categories, then providers and tools. Contracts describe the inputs and outputs for execution through firecrawl_scrape. nextTool identifies further discovery or pagination. Discovery does not execute providers. Use firecrawl_search when you also need web results.

Start a multi-page crawl at a website URL, poll it to a terminal state, and return the final status and collected data. Scope can be bounded with include/exclude paths, depth, page limit, subdomain/external-link controls, sitemap handling, delay, and scrape options. Crawl results can be large; use conservative limits when full-site coverage is unnecessary. Webhooks and interactive scrape actions are unavailable in safe mode. Returns the crawl ID, status, and page data.

Retrieve the current status, progress, and available results for an existing crawl ID. This only reads Firecrawl job state and does not start or modify the crawl.

Run web research that returns structured data when the URLs are not known or the answer spans several sites. Describe the fields you need in `prompt`, optionally pass a JSON `schema` and seed `urls`, and the research agent searches, navigates, reads pages, and returns JSON assembled across sources. Use it to research an entity plus its fields (founders, pricing, contact details), to build lists and datasets (companies, people, products, jobs, papers), and for pages that need navigation or interaction to reach the data. Optional `effort` sets the reasoning budget, `maxCredits` caps spend, and `strictConstrainToURLs` keeps the agent to the supplied `urls`. This call returns only a job ID, not the research result. Read the job with `firecrawl_agent_status` until it reaches `completed` or `failed`; a typical research run takes one to three minutes. For one known URL use `firecrawl_scrape` (with formats: ["json"] for structured output); for a plain lookup that a results page answers, use `firecrawl_search`. The job also returns a `threadId`. To continue that thread, pass it with a follow-up `prompt`; omitted `mode`, `urls`, `schema` and exchange settings carry over from the previous turn. The agent only calls Alexandria providers whose terms the team has accepted. `exchange.skippedProviders` lists gated providers. With `exchange.onTermsRequired` "ask", a terms `pendingApproval` and `exchange.requiresAction` carry the `approvalId` and provider requirements. Read terms/show through `firecrawl_scrape`. An organization admin must accept terms in the Firecrawl dashboard at the provider URL or https://www.firecrawl.dev/app/settings?tab=data-sources. Ignore any terms/accept call in the API response. Only after the admin confirms acceptance, resume with the same `threadId` and `exchange.approve: {approvalId}`; this does not accept terms. To decline, use `exchange.decline: {approvalId}`. Never infer acceptance from a data request.

Retrieve progress or final results for a `firecrawl_agent` job ID. A `processing` response is non-terminal and does not contain the final research result. Check again after 15–30 seconds until the status is `completed` or `failed`; a typical research run takes one to three minutes and complex jobs can take longer. If the job cannot finish within the task's available time, use `firecrawl_search` and `firecrawl_scrape` to complete the requested output. Returns job status, progress information, and result data when completed.

Open or reuse a live browser session to navigate a page, click controls, fill fields, or run browser code. Provide either `url` or `scrapeId`, and either a natural-language `prompt` or executable `code`; code can run as Bash, Python, or Node with a bounded timeout. This acts on the live site, so actions such as form submission can create persistent external side effects. Returns execution output, stdout/stderr, exit status, and session viewing URLs.

Stop the live interact session associated with a `scrapeId` and release its resources. Returns a success confirmation.

Parse one supported document into markdown, HTML, links, summary, targeted answers, or JSON matching a schema. Supported inputs include common HTML, PDF, Word, RTF, OpenDocument, and spreadsheet files; PDF parsing can be bounded with `pdfOptions.maxPages`. Local MCP reads `filePath` from the server filesystem. Hosted MCP uses two calls: first provide `filePath` to receive upload instructions, upload locally, then call again with the returned `uploadRef`; do not send both fields together. Remote web URLs belong in `firecrawl_scrape`. Set `redactPII` to request redaction of personally identifiable information in the returned content. `zeroDataRetention` requires an eligible authenticated account; omit it for anonymous keyless use. Returns upload instructions for hosted phase one or parsed document content for the final call. Authenticated final responses can include a `data.metadata.scrapeId` for optional parse feedback.

Create a recurring scrape, crawl, or search monitor that compares each check with its retained predecessor. The simple form accepts `page`/`pages` or `queries` plus a plain-language `goal`; the advanced `body` form controls targets, schedule, change-tracking formats, judging, retention, webhook, and notifications. In the simple form, a `goal` is required. If `queries` contains one or more non-empty values and is supplied with `page`/`pages`, `queries` create the search target and page targets are ignored. A monitor schedules future network checks and can send configured email or webhook notifications. Returns the created monitor.

List monitors for the authenticated account with optional pagination controls. Returns one page of monitor records and pagination metadata.

Retrieve one monitor by ID, including its configuration and current state. This does not run or modify the monitor.

Patch an existing monitor by ID. The body can change its name, active/paused status, schedule, targets, goal, judging, webhook, notifications, or retention; these changes affect future scheduled checks. Returns the updated monitor.

Permanently delete a monitor by ID and stop its future schedule. This operation cannot be undone and returns deletion status.

Queue an immediate check for a monitor outside its normal schedule. This starts network work for the monitor's configured targets and returns the queued check.

List historical checks for a monitor, optionally filtered by status and bounded by a result limit. Returns one page of check summaries and pagination metadata.

Retrieve one monitor check and its page-level results, optionally filtered by page status. Pages report `same`, `new`, `changed`, `removed`, or `error`; configured goal judging can add a meaningful-change decision. Markdown tracking returns a unified text diff, JSON tracking returns field paths with previous/current values and a current snapshot, and mixed tracking returns both. Returns one page of results plus a `next` URL when more pages exist.

Search paper metadata and abstracts with a natural-language query across the indexed corpus, which spans biomedical, life-science, and clinical literature (PubMed, bioRxiv, medRxiv) alongside arXiv and other scientific sources. Optional author, category, date, and sort options constrain or reorder results. Several distinct framings of the same question surface different papers than a single query does. Returns ranked papers with canonical IDs, titles, authors, and abstracts.

Retrieve canonical metadata for one paper ID, such as an arXiv, PMC, PMID, or DOI identifier. Returns the title, abstract, authors, categories, source IDs, and dates as markdown.

Find citation-graph candidates from one to ten `seed_ids`; the first ID is the primary seed and later IDs are anchors. `mode` defaults to `similar` (co-citation/bibliographic coupling); `citers` returns papers citing a seed and `references` papers cited by a seed. `intent` ranks candidates. Returns ranked candidates and the evaluated pool size.

Retrieve in-body passages from one paper that are relevant to a specific question. Full text is available only for indexed papers; `k` controls the number of passages. Returns matching passages or a notice when full text is unavailable.

Search an index of public repositories, GitHub issues, merged pull requests, repository READMEs, and code documentation for programming questions that need external documentation or upstream evidence. Returns ranked results with an ID, source type, URL, title, and the matched passages in markdown.

Search US federal, state, and local statutes, regulations, codes, court opinions, and other government publications for legal or regulatory questions that need primary text. Clarify a missing jurisdiction or ambiguous as-of date before identifying the governing rule. Optional filters narrow results by jurisdiction, court, court type, cited law or case, and site. Filters keep only pages the index has labeled, so a filtered search can leave out relevant pages, and no results does not mean none exist. Court labels are most complete for federal courts and case-law collections. Returns ranked results with a position, title, URL, and matched snippet. Results may include historical versions, bill text, agency guidance, and third-party reproductions. Verify the issuer, jurisdiction, enacted/effective status, and version before citing current governing law. Read a known URL or additional page context with firecrawl_scrape. Use general search for non-US questions or missing coverage; no hits does not establish that no applicable law exists.

Get the authenticated Firecrawl team's current credit balance or its credit usage over the last 90 days. Use the current view to answer how many credits remain or to check billing-period boundaries. Use the historical view to answer what credits were spent on and when. The current view returns `remainingCredits`, `planCredits`, `billingPeriodStart`, and `billingPeriodEnd`. Extra purchased or granted credits can make `remainingCredits` greater than `planCredits`. Billing-period dates may be null when the upstream billing provider has no period metadata. The historical view returns periods sorted by start date with `startDate`, `endDate`, and `creditsUsed`, one per calendar month, or one per UTC day with `bin: "day"`. Set `byEndpoint` to split each period by the endpoint that used the credits (`endpoint`: scrape, search, crawl, map, agent, ...), or `byApiKey` to split it by API key (`apiKey`). The endpoint split also counts search usage, which the unsplit and per-key figures leave out, so its rows can add up to more. The newest period's `endDate` can be null. A `notice` means the requested split or daily periods are not available yet and the periods are monthly figures. This tool cannot attribute credits to individual jobs, URLs, or MCP clients. Each tool's own result reports what that call used (`creditsUsed`, or `metadata.creditsUsed` on scrapes), and the Usage and Logs pages of the Firecrawl dashboard list usage per job.

概览

什么是 Firecrawl MCP Server?

Firecrawl MCP 是 Firecrawl 官方推出的 Model Context Protocol 服务器,让 AI 智能体可以在 Claude、Cursor、VS Code 等客户端中搜索网页、把页面抓取为干净的 Markdown 或 JSON、爬取和映射站点,并与实时页面交互。

你可以直接使用托管端点 https://mcp.firecrawl.dev/v2/mcp,无需安装;也可以通过 npx 或 Docker 在本地运行开源的 firecrawl-mcp 包。本地服务器同样支持自托管的 Firecrawl 实例。

如何使用 Firecrawl MCP Server?

前提条件: 要使用完整工具集,需要从 https://www.firecrawl.dev/app/api-keys 获取 Firecrawl API 密钥。只有本地 npx 安装才需要 Node.js。

托管、免密钥: 将客户端指向 https://mcp.firecrawl.dev/v2/mcp。没有密钥时,firecrawl_scrape、firecrawl_search 和 firecrawl_parse 可在限流条件下使用;爬取、映射和 agent 类工具需要密钥。

托管、登录: 使用 https://mcp.firecrawl.dev/v2/mcp-oauth,并在客户端中完成 OAuth 流程。这个 URL 是 MCP 端点,不是要在浏览器中打开的网页。

托管、API 密钥: URL 保持为 https://mcp.firecrawl.dev/v2/mcp,并发送请求头 Authorization: Bearer <FIRECRAWL_API_KEY>。README 明确提醒:不要把密钥放进 URL,也不要在与智能体的对话中粘贴密钥。

本地:

env FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp

Claude Code

# Hosted with OAuth sign-in
claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp-oauth

# Local server
claude mcp add firecrawl -e FIRECRAWL_API_KEY=fc-YOUR_API_KEY -- npx -y firecrawl-mcp

Claude Desktop

在 claude_desktop_config.json 中添加:

{
  "mcpServers": {
    "mcp-server-firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Cursor

打开 Cursor Settings > Features > MCP Servers,点击 + Add new global MCP server,然后输入:

{
  "mcpServers": {
    "firecrawl-mcp": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR-API-KEY"
      }
    }
  }
}

VS Code

在 .vscode/mcp.json 中配置,运行时提示输入密钥而不是明文保存:

{
  "inputs": [
    { "type": "promptString", "id": "apiKey", "description": "Firecrawl API Key", "password": true }
  ],
  "servers": {
    "firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": { "FIRECRAWL_API_KEY": "${input:apiKey}" }
    }
  }
}

Windsurf

将同样的 mcpServers 配置块(服务器名为 mcp-server-firecrawl)添加到 ./codeium/windsurf/model_config.json。

如需自托管,把 FIRECRAWL_API_URL 设置为你的实例地址。设置 HTTP_STREAMABLE_SERVER=true 可让本地服务器通过 Streamable HTTP 在 http://localhost:3000/mcp 提供服务。Docker 镜像发布为 ghcr.io/firecrawl/firecrawl-mcp-server。

Firecrawl MCP Server 的主要功能

  • 网页搜索可返回完整页面内容,另有面向编程智能体的搜索索引,覆盖 GitHub issue、已合并的 pull request、README 和文档
  • 将任意 URL 抓取为 Markdown、HTML、链接、截图,或符合你所提供 schema 的 JSON
  • 爬取和映射整站,支持路径过滤、深度和页数限制
  • 实时浏览器交互:在读取页面前进行点击、输入和导航
  • 异步研究智能体、文档解析(PDF、Word、电子表格),以及带差异对比的定期页面监控
  • 额度用量查询、自动重试和限流处理
  • 支持托管(免密钥、OAuth 或 API 密钥)或通过 npx/Docker 本地运行,包括对接自托管 Firecrawl

Firecrawl MCP Server 工具

在完整配置和默认设置下,服务器会列出 27 个工具;免密钥的托管端点只提供 firecrawl_scrape、firecrawl_search 和 firecrawl_parse。

  • 抓取与解析: firecrawl_scrape 读取一个已知 URL;firecrawl_parse 转换 PDF 等文档。结构化提取通过 firecrawl_scrape 加 formats: ["json"] 完成(firecrawl_extract 已弃用)。
  • 发现: firecrawl_search 从查询词出发;firecrawl_map 列出站点 URL 但不抓取内容。
  • 爬取: firecrawl_crawl 收集多个页面;firecrawl_check_crawl_status 查看任务进度。
  • 交互: firecrawl_interact 打开实时浏览器会话;firecrawl_interact_stop 关闭会话。
  • 研究: firecrawl_agent 和 firecrawl_agent_status 运行多来源研究任务;firecrawl_research_* 工具可检索 PubMed、bioRxiv、medRxiv 和 arXiv 中的论文。
  • 监控: firecrawl_monitor_* 工具安排定期检查并报告变化。
  • 账户与反馈: firecrawl_credit_usage、firecrawl_search_feedback 和 firecrawl_feedback。

Firecrawl MCP Server 的使用场景

  • 在编程会话中拉取最新文档:“Scrape the Next.js caching docs and summarize what changed.”
  • 带出处的调研:“Search for recent benchmarks of vector databases and give me the top five with links.”
  • 结构化提取:“Get the plan names and prices from this pricing page as JSON.”
  • 站点审查:“Map example.com and list every page under /blog.”
  • 需要交互才能显示的页面:“Open this page, click Load more, then read the results.”
  • 变化追踪:为竞争对手的定价页设置监控,页面变化时收到提醒。

Firecrawl MCP Server 与替代方案对比

Exa MCP Server 和 Tavily MCP 专注于面向智能体的搜索,从自有索引返回内容;Firecrawl MCP 则以抓取、爬取和与你指定的页面交互为核心。Apify Actors MCP 提供针对特定网站的现成爬虫。如果只需要简单的浏览器自动化而不依赖抓取 API,Playwright MCP 会在本地运行浏览器。

Firecrawl MCP Server 常见问题排查

  • 只出现三个工具: 你正在使用免密钥的托管端点。通过 /v2/mcp-oauth 登录或添加 API 密钥,即可解锁 crawl、map、agent 等其余工具。
  • 限流错误: 服务器会把 Firecrawl 的限流错误返回给客户端。登录或使用 API 密钥可获得更高限额;进行爬取时也可以调低 limit。
  • Windows 上 Cursor 命令出错: 尝试 cmd /c "set FIRECRAWL_API_KEY=your-api-key && npx -y firecrawl-mcp"。
  • 在浏览器中打开 OAuth URL 没有效果: 它是服务器 URL;请在 MCP 客户端中添加,并使用客户端的账号连接流程。
  • 工具数量超过客户端上限: 设置 FIRECRAWL_NO_SEARCH_FEEDBACK=1 和/或 FIRECRAWL_NO_ENDPOINT_FEEDBACK=1,或使用仅搜索端点 https://mcp.firecrawl.dev/v2/mcp-search。

常见问题

1

Firecrawl MCP 是官方的吗?

是的。该服务器由 Firecrawl 在 `firecrawl/firecrawl-mcp-server` 仓库(原 `mendableai`)中维护,并感谢 @vrknetha 最初的社区实现。

2

Firecrawl MCP 服务器免费吗?

代码采用 MIT 许可证,托管的免密钥档位无需账号即可在限流条件下使用 scrape、search 和 parse。其他工具会消耗你 Firecrawl 套餐的额度;当前价格请查看 firecrawl.dev。

3

需要哪种 API 密钥?

需要从 `https://www.firecrawl.dev/app/api-keys` 获取的 Firecrawl API 密钥(以 `fc-` 开头)。本地使用时设置为 `FIRECRAWL_API_KEY`,托管使用时作为 Bearer token 发送。托管模式也可以改用 OAuth 登录。

4

Firecrawl MCP 支持 Claude Code、Cursor 和 VS Code 吗?

支持。README 提供了 Cursor、VS Code、Windsurf 和 Claude Desktop 的配置,Claude Code 可以添加托管 URL 或本地 npx 命令。

5

应该用远程服务器还是本地服务器?

托管服务器无需安装并支持 OAuth。如果你想通过 `FIRECRAWL_API_URL` 对接自托管 Firecrawl、解析本地磁盘上的文件,或用 Docker 控制运行环境,就选择本地运行。

6

让智能体使用 firecrawl_interact 安全吗?

`firecrawl_interact` 会作用于真实网站,提交表单等操作可能产生实际影响。批准这类工具调用前请先检查。

7

可以不依赖 Firecrawl 云服务运行 Firecrawl MCP 吗?

可以。将 `FIRECRAWL_API_URL` 设置为自托管的 Firecrawl 实例;只有当你的实例要求认证时才需要 API 密钥。

评论

浏览器自动化 分类下的更多 MCP 服务器