How to Scrape Goofish (Xianyu / 闲鱼) in 2026 Without +86 Login or Cookies (Python SDK + MCP)
Goofish (闲鱼 / Xianyu) is Alibaba's flagship C2C recommerce marketplace — processing over 100M+ monthly active buyers across vintage mirrorless cameras, PC GPUs, retro handhelds, road bikes, luxury handbags, and anime/ACG collectibles (guzi / 谷子). Because individual Chinese sellers price used gear fo

r fast domestic turnover, street prices on Goofish routinely sit 20%–45% below eBay, Reverb, Mercari JP, or Chrono24. Yet if you have ever tried building a Python deal monitor, price-comp dataset, or AI shopping agent for Goofish from outside mainland China, you hit four immediate engineering walls: Ephemeral Request Signing & Session Tokens: Goofish's web and mobile endpoints require short-lived cryptographic request signatures paired with rotating session tokens that expire rapidly. Alipay / +86 Real-Name Login Walls: Unauthenticated keyword search and seller-profile requests on standard web views immediately hit a mandatory Taobao/Alipay QR login gate requiring mainland +86 SMS or passport KYC verification. Aggressive Datacenter Anti-Bot Filtering: Standard cloud IPs (AWS, GCP, GitHub Actions, Hetzner) are immediately blocked or challenged with interactive verification sliders.
Opaque m.tb.cn Share Snippets & Slang: Sellers share links as raw mobile text (【闲鱼】https://m.tb.cn/h.xxxx 点击链接直接打开), and listings use domain-specific Chinese shorthand (箱说全 = full box & papers, 无拆无修 = unopened/unrepaired, 个人闲置 = personal seller, 吧唧 = anime tin badge) that literal English keyword searches miss completely. In this guide, we will walk through how to query live Goofish listings, resolve m.tb.cn share links, compute CNY-to-USD markdowns, generate 1-click Superbuy / AllChinaBuy proxy checkout links, and audit seller Zhima Credit (芝麻信用) & buyer reviews using the open-source goofish-xianyu-api Python SDK and goofish-xianyu-mcp Model Context Protocol server — with zero Taobao/Alipay cookies. Instead of maintaining brittle local Playwright browsers, +86 SIM cards, and residential proxy pools, goofish-xianyu-api (zero third-party Python dependencies) and goofish-xianyu-mcp expose a typed client + local helpers (offline English-to-Chinese Xianyu slang glossary, itemId watchlist state deduplication, CNY-to-USD markdown math, and Zhima Credit scoring) backed by cloud HTTP endpoints: Module / Tool Package Key Capabilities GoofishClient.search_items / search_many goofish-xianyu-api Single & multi-keyword search, cross-keyword itemId dedup, auto-pagination, only_new state file GoofishClient.get_item_detail / get_seller_* goofish-xianyu-api m.tb.cn share-link resolution, seller Zhima Credit & KYC badges, onsale/sold comps, buyer reviews search_goofish_items & 4 Seller/Item MCP Tools goofish-xianyu-mcp Native stdio MCP server for Claude Desktop, Cursor, Windsurf, and VS Code Before writing code, it helps to separate what public guest extraction reliably delivers from what requires a personal mobile app session: Works Reliably Without Cookies: Live keyword search with CNY price_min / price_max, personal_only (个人闲置), free_shipping (包邮), sort order (newest, price_asc, priceDrop, credit), and city region filters. Multi-page Standby auto-pagination (30 items/page up to 150 items synchronously, or up to 1,500 items via batch fallback).
Full item detail resolution (description, HD image gallery, specs dictionary, wants count, city, condition tag) from numeric IDs, goofish.com URLs, or raw m.tb.cn mobile share text. Seller Zhima Credit (芝麻信用) tiers (极好, 优秀, 良好, 中等, 较差), biometric KYC verification badges (实人认证, 实名认证), active onsale and historical sold catalogs, and paginated buyer feedback reviews (all, negative_only, with_photos). Upstream Platform Boundaries: Private buyer-seller DM chat messages, unmasked seller phone/Alipay identifiers, and direct order placement require an authenticated personal Goofish app account (or a cross-border purchasing agent such as Superbuy or AllChinaBuy). Both goofish-xianyu-api and goofish-xianyu-mcp auto-detect credentials from your environment (APIFY_TOKEN, UNITBYTES_API_KEY, or RAPIDAPI_KEY; see package READMEs for token setup): export APIFY_TOKEN="apify_api_YOUR_TOKEN" # Also supports UNITBYTES_API_KEY or RAPIDAPI_KEY automatically goofish-xianyu-api) Install the typed Python SDK (Python 3.10+): pip install goofish-xianyu-api goofish-xianyu-api ships with a curated offline English-to-Simplified-Chinese Xianyu glossary (translate_keyword, zero LLM calls), cross-keyword itemId deduplication (search_many), atomic local state persistence (only_new=True), and proxy-agent URL builders (build_proxy_agent_links): from goofish_xianyu_api import ( GoofishClient, analyze_resale_pricing, build_proxy_agent_links, translate_keyword, ) # 1.
Offline glossary preserves model numbers while translating Xianyu slang print(translate_keyword("Sony A7C2 mirrorless camera box and papers")) # -> "索尼 A7C2 微单相机 箱说全" print(translate_keyword("Genshin Impact badge unopened")) # -> "原神 吧唧 全新未拆封" client = GoofishClient() # Reads APIFY_TOKEN from environment # 2. Multi-keyword watchlist sniper (auto-paginates >30 items & deduplicates itemIds) unseen_deals = client.search_many( [ "Sony A7C2 mirrorless", "Fujifilm X100VI box and papers", "Nintendo Switch OLED", ], max_items=45, translate_keywords=True, personal_only=True, # Exclude commercial storefronts (个人闲置) free_shipping=True, price_min=900, price_max=13500, sort_by="newest", only_new=True, # Persist seen item IDs so cron runs emit only new drops state_file="watchlist_seen.json", ) for item in unseen_deals[:5]: pricing = analyze_resale_pricing(item, cny_to_usd=0.138) links = build_proxy_agent_links(item) print( f"[{item.get('matchedKeyword')}] {item['title']}\n" f" Price: ¥{pricing.price_cny} (${pricing.price_usd}) | " f"Markdown: -{pricing.discount_pct}% | Condition: {pricing.condition_normalized}\n" f" Superbuy: {links.superbuy}\n" f" AllChinaBuy: {links.allchinabuy}" ) Cheap electronics or luxury pieces on C2C marketplaces require seller vetting before you submit an agent order. You can score a seller's Zhima Credit (芝麻信用), biometric KYC verification, historical sold count, and negative dispute reviews programmatically: from goofish_xianyu_api import GoofishClient, assess_seller_trust client = GoofishClient() # Resolve a raw mobile share link first if you only have pasted text item = client.get_item_detail("【闲鱼】https://m.tb.cn/h.5xyz123 点击链接直接打开") seller_id = item.get("sellerId") or "2214891234567" # Fetch seller profile + compute objective 0-100 trust assessment profile = client.get_seller_profile(seller_id) trust = assess_seller_trust(profile) print( f"Seller Trust: {trust.trust_score}/100 | " f"Zhima Tier: {trust.zhima_tier} | " f"KYC Verified: {trust.is_kyc_verified} | " f"Risk: {trust.risk_level}" ) print("Signals:", trust.signals) # Inspect historical sold comps to check if seller is a genuine hobbyist or grey-market dealer sold_history = client.get_seller_items(seller_id, status="sold", max_items=20) print(f"Sampled {len(sold_history)} historical sold items") # Pull negative buyer feedback specifically to catch refurbishment / dispute red flags disputes = client.get_seller_reviews( seller_id, review_filter="negative_only", max_reviews=20, ) for rev in disputes: print("Dispute review:", rev.get("content")) json, jsonl, csv) Installing goofish-xianyu-api also registers the goofish-xianyu (and goofish-scrape) CLI for one-liner CSV exports and cron scripts: # Multi-keyword translated watchlist -> JSONL (only newly listed items) goofish-xianyu --format jsonl search \ --keywords "Sony A7C2 mirrorless" "Pokemon card unopened" \ --translate \ --personal-only \ --max-items 40 \ --only-new ./seen_items.json # Export personal-seller deals with USD markdowns & Superbuy columns to CSV goofish-xianyu --format csv --output camera_deals.csv search "理光 GR3x 箱说全" --personal-only --max-items 30 If you prefer researching secondhand comps conversationally inside Claude Desktop, Cursor, Windsurf, or VS Code Copilot, install the zero-dependency goofish-xianyu-mcp server in one command: # Install into Cursor (~/.cursor/mcp.json) npx -y goofish-xianyu-mcp --install cursor --token apify_api_YOUR_TOKEN # Or install into Claude Desktop npx -y goofish-xianyu-mcp --install claude --token apify_api_YOUR_TOKEN Or add it manually to mcpServers: { "mcpServers": { "goofish-xianyu": { "command": "npx", "args": ["-y", "goofish-xianyu-mcp"], "env": { "APIFY_TOKEN": "apify_api_YOUR_TOKEN" } } } } Once connected, your AI assistant gains 5 live tools (search_goofish_items, get_goofish_item_detail, get_goofish_seller_profile, get_goofish_seller_reviews, and get_goofish_seller_items) — including multi-keyword keywords deduplication, auto-pagination up to 100 listings, offline translateKeywords slang conversion, and ready-to-click Superbuy/AllChinaBuy links. Try prompting Claude or Cursor: "Search Goofish personal sellers (personalOnly=true, translateKeywords=true) across keywords: ['Sony A7C2 mirrorless', 'Fujifilm X100VI box and papers'] up to 40 items between ¥7,500 and ¥12,000.
Filter out local-pickup-only posts, audit the top 2 sellers' Zhima Credit and negative buyer reviews, and give me Superbuy links." Python SDK (PyPI): pypi.org/project/goofish-xianyu-api (pip install goofish-xianyu-api) MCP Server (npm): npmjs.com/package/goofish-xianyu-mcp (npx -y goofish-xianyu-mcp) Technical Documentation: unitbytes.com/actors/goofish-xianyu-search-scraper Disclaimer: Independent integration powered by UnitBytes; not affiliated with or endorsed by Alibaba Group, 1688.com, or Goofish/Xianyu.
Key Takeaways
- •Yet if you have ever tried building a Python deal monitor, price-comp dataset, or AI shopping agent for Goofish from outside mainland China, you hit four immediate engineering walls: Ephemeral Request Signing & Session Tokens: Goofish's web and mobile endpoints require short-lived cryptographic request signatures paired with rotating session tokens that expire rapidly.
- •This story was reported by Dev.to, covering developments in the dev space.
- •AI advancements continue to reshape industries — read the full article on Dev.to for complete coverage.
📖 Continue reading the full article:
Read Full Article on Dev.to →


