AnySearch

Unified real-time search engine skill for AI agents.

该来源不提供完整文件导出（国内平台多为平台内托管），仅存元数据与原链

接入信息

传输形态: http
鉴权方式: 需 API Key（需要配置：AnySearch API Key）
端点: https://anysearch.run.tools

该服务需要凭证，请按官方文档申请后替换占位符

{
  "mcpServers": {
    "AnySearch": {
      "headers": {
        "Authorization": "Bearer \u003cYOUR_KEY\u003e"
      },
      "url": "https://anysearch.run.tools"
    }
  }
}

能力清单

工具	说明
batch_search	Run multiple searches in parallel and return all results merged into one response. ## When to use Use batch_search instead of multiple sequential search calls when you have 2–5 independent queries. This saves context window space by returning all results in a single tool call. ## Constraints - Maximum 5 queries per call - Each query item has the same structure as the search tool parameters - Queries run in parallel; a single query failure does not block others ## Example Instead of: search(query=A) → search(query=B) → search(query=C) Use: batch_search(queries=[{query:A,...}, {query:B,...}, {query:C,...}])
extract	Fetch a URL and return its full content as clean Markdown. ## When to use — call extract after search whenever: - The search snippet is too short or truncated to answer the question - User asks to 'read', 'open', 'summarize', or 'get details from' a specific URL - You need to verify a specific claim, statistic, or fact from the original source - The result points to a full article, report, documentation page, or paper worth reading in full - The answer requires data visible only in the page body (tables, sections, code blocks not captured in snippet) - User provides a URL directly and asks about its content ## When NOT to use - The search snippet already contains a complete, sufficient answer - You only need the URL or title (not the page body) ## Constraints - url must start with http:// or https:// - Only HTML pages are supported; PDF/binary files will return an error - Content is truncated at 50,000 characters
list_domains	Call this before search to get the sub-domain catalog and MANDATORY query format rules for a given domain. ## When to call — pick the domain that matches what the user is asking about: finance → stocks, ticker, ETF, forex, exchange rate, currency, commodities, oil/gold price, crypto, Bitcoin, earnings, SEC filing, IPO, bond yield, financial news academic → paper, research, journal, thesis, citation, DOI, abstract, peer review, arxiv, pubmed, scholar, literature, study, dataset ip → patent, prior art, invention, CPC, IPC, EPO, WIPO, assignee, inventor, trademark legal → law, statute, regulation, case, ruling, judgment, court, legislation, compliance, civil code, criminal code travel → flight, airline, departure, arrival, delay, airport, IATA, POI, attraction, hotel, itinerary, travel guide, visa gaming → game, Steam, price, discount, esports, player stats, rank, champion, LOL, DOTA, CS2 security → malware, virus, CVE, vulnerability, IP reputation, threat, IOC, hash, VirusTotal, OSINT, phishing, ransomware geo → address, coordinates, geocode, POI, nearby, restaurant, walkability, transit score environment → weather, forecast, AQI, air quality, PM2.5, satellite, NDVI, carbon, emission, agriculture energy → electricity price, power grid, oil price, gas price, energy market business → job, hiring, salary, company contact, B2B, lead, recruiter, HR code → library docs, API reference, npm, pip, cargo, code snippet, function, repo search health → clinical trial, diagnosis, drug, symptom, medical literature, WHO stats, psychology education → course, lecture, textbook, tutorial video, MOOC, open courseware tech → product specs, barcode, HackerNews, ProductHunt, tech review ecommerce → price comparison, Walmart, shopping, product search film → movie, TV show, anime, torrent, streaming music → album, artist, lyrics, music torrent fashion → cosmetics ingredients, beauty, trend, streetwear release home → recipe, repair guide, food safety, walkability, appliance religion → bible, quran, torah, buddhist texts, manuscripts ugc → Bilibili video, YouTube video ## Returns Markdown table filtered to the specified domain: sub_domain \| description \| query_format \| zone ## CRITICAL: How to use results - sub_domain is the PRIMARY routing key — always pass it to search - query_format column is MANDATORY — wrong format = wrong data source = wrong results Hard constraints examples: finance.us_stock requires Stock:/Forex:/News:/Commodities: prefix; security.noise requires single IPv4; geo.weather requires city name or lat,lon - If multiple sub_domains match different aspects, make PARALLEL search calls — one per sub_domain - zone=CN → set zone="cn" in search; zone=ALL/US/EU → omit zone ## Cache Rule — NEVER repeat list_domains for the same domain Once you have called list_domains for a domain in this conversation, the result is valid for the ENTIRE session. Do NOT call list_domains again for the same domain — reuse the sub_domain and query format you already received. If you need info for multiple domains, pass them all in the domains array in a single call.
search	Execute a search and return ranked Markdown results (title, URL, snippet). ## Two modes ### Mode 1 — General web search (no list_domains needed) Omit domain and sub_domain entirely. Use when the query is open-ended and does not target a specific structured data source. Example: search(query="what is quantum computing") ### Mode 2 — Vertical search (call list_domains first) Use when the query targets a specific domain: stocks, patents, flights, CVEs, weather, academic papers, etc. Steps: 1. Call list_domains to get the sub_domain and mandatory query format for the target domain. 2. Pass domain + sub_domain from list_domains output. Never guess them. 3. Format query exactly as specified in the query_format column — wrong format = wrong results. ## Decision rule — which mode to use Use Mode 2 (vertical) when ANY of these apply: - Query involves a ticker, DOI, CVE, IATA code, patent number, address, or other structured identifier - Query targets a specific vertical: finance, legal, academic, travel, security, geo, environment, etc. - User asks for real-time or specialized data (stock price, weather, flight status, drug info, etc.) Use Mode 1 (general) when the query is purely conversational or open-ended with no structured lookup. ## After getting results — when to call extract Search returns titles + snippets only. Call extract when: - The snippet is truncated or insufficient to answer the question - User asks to read, summarize, or get details from a specific URL - You need to verify a claim or fact from the source page - The answer requires data only visible in the page body (tables, sections not in snippet) ## Query decomposition One intent per search call. For 2–5 independent queries use batch_search instead. WRONG: search(query="AAPL price and earnings and analyst rating") RIGHT: batch_search(queries=[{query:"AAPL price",...}, {query:"AAPL earnings",...}])

纠错与举报（发现条目失效、署名有误或涉及侵权？）

提交举报 / 纠错

侵权举报经核验成立后，我们会即时下线该条目并删除已存的内容副本。