New: Crawl. Crawl a whole website from a single start URL and get every page back as Markdown or HTML. Geekflare follows links (and sitemaps) from the start URL, fetches each page once, and handles JavaScript rendering, retries, and proxies for you.
POST /crawl/syncwaits for the crawl and returns up to 500 pages in one response. If it runs out of time, you get the pages finished so far, marked as partial.POST /crawlstarts a background job for up to 10,000 pages and returns a job ID right away. Track it withGET /crawl/{id}, see how a page was reached withGET /crawl/{id}/trace, receive webhooks as it runs, and download the results as a ZIP when it finishes.- Control the scope with
maxDepth,maxPages,allowPatterns,denyPatterns, andincludeExternal. SetrenderJStofalsefor static sites, and useproxyModeorproxyCountrywhen pages need a proxy.
proxyMode now defaults to auto for Web Scraping. Requests that don’t set proxyMode now try without a proxy first and automatically retry through a proxy if the site blocks the request. Previously the default was false (never use a proxy). Set proxyMode: "false" to keep the old behavior, or "true" to always use a proxy. proxyCountry still works on its own. See Proxy Routing.Geekflare is now available in more places. Use Geekflare for search, page reading, screenshots, and website checks from the tools you already work in. See the full list on the integrations page and in Integrations & SDKs.
- AI chat: install Geekflare from the Claude directory or add the ChatGPT plugin. Both sign in with your Geekflare account, so no API key is needed.
- AI frameworks: LangChain, LlamaIndex, and the Vercel AI SDK each have a Geekflare tools package.
- Workflow tools: the Dify plugin is on the Dify Marketplace, and Geekflare is also available for Zapier, Make, n8n, viaSocket, and Flozic.
- Developer tools: the Raycast extension, plus MCP setup guides for Claude Code, Cursor, Codex, Gemini CLI, Kiro, and more. The official agent skill teaches coding agents when to use each Geekflare tool.
https://mcp.geekflare.com/mcp now supports sign-in with your Geekflare account, in addition to API-key and npx setups. Sign-in has been tested in Claude and Cursor. See Claude Desktop, Claude Code, and Cursor.New:
device for Search. Search now accepts device (desktop or mobile, default desktop) to get results as they appear on a mobile device. Works in standard search and SERP mode.AI Overviews in SERP. SERP responses now include data.ai_overview when Google shows an AI Overview for the query — the overview as text and HTML, plus the sources it cites. See AI Overviews.New: SERP mode for Search. Set
serp: true on Search to get the full Google search results page — organic results, related searches, People Also Ask, and pagination — instead of the standard response. Costs the same as a standard Search request (2 credits).City-level targeting. In SERP mode, pass city (for example "London,England,United Kingdom") to localize results to a specific city, optionally alongside location. See City Targeting and the supported cities.Meta Scraping is now 1 credit. Reduced from 2 credits per request. See Credit Mapping for the full pricing table.
New:
detectServices for Port Scanner. Port Scanner now supports detectServices: true to run service/version detection on the ports found open, identifying what’s actually running on each one. It’s best-effort and slower than a base scan since it probes each open port individually. If detection fails, the port list is still returned without service info.New: Brand endpoint. Get structured brand information for any domain — name, logos, colors, fonts, social profiles, links, and page metadata — in a single request.Two modes are available via
mode:standard(default) — brand identity and design tokens: name, colors, fonts, favicon, links, social profiles, and page metadata.enriched— everything instandard, plus LLM-synthesized company intelligence: industry, founding year, headcount and revenue ranges, target audience, brand voice, and use cases.
TLS Scan now checks for legacy SSL versions. TLS Scan now detects whether a server still supports the deprecated and insecure SSLv2 and SSLv3 protocols, so you can catch outdated TLS configurations.New extraction templates for Web Scraping. Web Scraping now supports
extractionMode: "template" with two ready-made templates: product (extract product info like title, price, and availability) and contact (extract contact info like emails, phone numbers, and addresses).Test Load Time from multiple countries. Pass
targetCountries (up to 3 ISO country codes) to test a URL’s reachability from those locations via proxy, alongside the default US server test, in a single request. Useful for confirming a site loads correctly for visitors in different regions, checking whether a site is blocked or filtered in a particular country, or diagnosing geo-specific outages.The response reports reachability per location, plus an overall summary. Costs a flat 8 credits per request regardless of how many locations (1–3) you test. See Load Time: Multi-Location Testing.IPv6 support. Ping and Port Scanner now accept IPv6 addresses and URLs in addition to IPv4, so you can test reachability and scan ports on IPv6-only or dual-stack hosts.
Simplified Web Scraping credits. All Web Scraping requests now cost a flat 1 credit, regardless of whether JavaScript rendering is used. The previous 2-credit “standard” tier has been removed. AI Extraction remains a +7 credit add-on (8 total).Automatic JavaScript rendering.
renderJS is now automatic by default — Geekflare fetches the page without a browser first, and only falls back to full JavaScript rendering if the page needs it, at no extra cost. Set renderJS: true or renderJS: false explicitly to force rendering on or off. See JavaScript Rendering.New proxyMode parameter. Web Scraping now supports proxyMode (false / auto / true, default false) for finer control over when a proxy is used:false— no proxy (default)auto— try without a proxy first, and automatically retry through a proxy if the request is blockedtrue— always route through a proxy
proxyCountry continues to work on its own, without proxyMode, to route through a proxy in a specific country. See Proxy Routing.