9 years of seller data · 40+ global marketplaces · 160+ data dimensions · Backed by Sorftime's AI data supply chain
AI-powered product research, competitor analysis, and keyword intelligence for Amazon, Walmart, TikTok Shop, Shopee, TEMU, 1688, Mercado Libre sellers — plus overseas social listening (Reddit/Quora).
Open source. Free trial. Works with any MCP-compatible AI agent — Claude Code, Codex, Cursor, OpenClaw, and more.
Stop clicking through dashboards. Just talk to your AI.
- 🐛 Fixed reverse parameter rewrite: the bridge was rewriting the current
keywordparameter ofproduct_searchto the retiredsearch_name(stale static schema + over-eager alias), which the server silently ignored and returned an unrelated default best-seller list.keywordnow passes through untouched; legacysearch_namecallers are still auto-corrected tokeyword. - 🧹 9 tool schemas regenerated from the live server (
product_search,product_search_from_history,product_trend,keyword_trend,category_report,*_search_from_name,ali1688_product_search) — plus a 30-day freshness guard on the local schema cache andaudit_static_schemas.py(static vs live drift gate) so this class of staleness can't hide again. - Tool counts unchanged (root 120 · amazon 48). Full changelog →
- 🔍
keyword_search_from_name(Amazon, 2 credits): describe a product in natural language and get matched trending keywords (20 per page). Live schema-verified; data coverage is backfilling (returns "No data available" for now — reported honestly). - 🏪
seller_detail(Amazon): full seller profile by name or ID — nationality/region, launch date, product count, feedback volume, province/city. Same backfill status. - ♻️ Quora trio re-mounted on the Amazon endpoint (server-side: removed 2026-10-27, restored 2026-11-06; social endpoint unchanged). Root endpoint: 118 → 120 tools; amazon endpoint: 43 → 48.
- Full changelog →
🆕 What's New (October 27, 2026) — v5.0–v5.2: Patent Suite, Dual-Site Registration & Dual-Channel Access
- ⚖️ Amazon patent suite (v5.2.0): 3 live-verified tools —
patent_search_from_image(design-patent image search, up to 100 hits with similarity scoring;query_modeeffectively required),patent_search_from_keyword(keyword / company / inventor / publication-number search, filterable by patent type, legal status, country, applicant, classification), andpatent_request(full detail by publication number — LOC/IPC/CPC classes, legal status, dates, patent figures). Data via Zhihuiya. - 📦 Richer product & keyword data (v5.2.0):
walmart_product_contentandshopee_product_content(complete image sets, attributes, variant details; Shopee adds shop coupons & location);product_trendgainsDailySalesVolume;keyword_search_resultsgainsposition_type(all / organic / ads only). Root endpoint: 113 → 118 tools. - 🌏 Official dual-site registration (v5.1): mainland open.sorftime.com (phone/WeChat, CNY) and international open-intl.sorftime.com (email/Google, PayPal/Stripe). Register on whichever site suits you — its Account-SK works across ALL platform endpoints with a shared quota; each site's pricing is shown on its own page.
- 🔌 Dual-channel data access (v5.0.0): Channel A — one-key configure the official
sorftime-<platform>native MCP services from the site's MCP 服务 page; this skill auto-detects them and serves as the knowledge layer (data queries need no local key; batch scripts still want a one-time key via install.py). Channel B — the built-in bridge viainstall.py(default). Auto-routed; no config juggling. - 🧱 Same-source platform packages: all 8 platform editions rebuild from this single repo (fix once, rebuild all); each ships a user README with a 3-step quickstart.
- Full changelog →
- 🌐 MCP split by platform (8 endpoints): Sorftime MCP now ships one endpoint per platform —
mcp.sorftime.com/{amazon|walmart|shopee|tiktok|temu|1688|mercado|social}— with the root endpoint kept as the all-platform compatible default (118 tools = exact union of the 8 endpoints at that time; now 120). One Account-SK works across all endpoints with a shared quota. Endpoint resolution:SORFTIME_MCP_URLenv > per-packageconfig/platform.json> root. - 📦 Platform packages via a single-source builder:
build/build_packages.pygenerates 8 self-contained skill packages (sorftime-amazon,sorftime-walmart,sorftime-shopee,sorftime-tiktok,sorftime-temu,sorftime-ali1688,sorftime-mercado,sorftime-social) from this repo — each with a platform-tailored hot layer, its own platform knowledge page, and a pre-wired endpoint. Fix once, rebuild all. - 🧠 Knowledge restructured for progressive disclosure: SKILL.md shrinks 58KB → 7KB; onboarding scripts, per-platform tool tables, gotchas, content QA and script guides move into
references/(8 newreferences/platforms/*.mdpages). The model reads what it needs, when it needs it. - 🎵 TikTok suite fully renamed server-side (breaking): the old 7 tools are retired; 18 new tools ship under the
*_request/*_search_from_nameconvention — now with product search (30+ filters) plus five new search families: shops, authors, brands, videos, hashtags. Same-args renames are auto-remapped; behavior-changed tools fail cleanly instead of mis-calling. - ✨ New Amazon capabilities (verified live):
common_usage_query(quota balance, zero-arg, every endpoint; briefly namedusage_queryearly on),category_low_price_search(Amazon Haul low-price-mall category intelligence),product_content(listing bullets/A+ content/buyer-focus teardown),seller_search+seller_search_from_name(40+ seller filters), and the Quora intel trio.seller_search/quora_*are live but still backfilling data coverage (verified no-data across 3 valid param sets each — reported honestly). - 🔧 New ops tooling:
scripts/probe_endpoints.py— one-command health check across all 9 endpoints with tool counts and credit balance,--matrixregenerates the tool matrix grouped by endpoint.healthcheck.pyis now endpoint-aware (works on any platform endpoint, detects rotated keys via pseudo-tools, shows remaining credits). - Full changelog →
1. Put the skill folder into your AI tool's skills directory (keep the folder intact):
| Tool | Directory |
|---|---|
| Claude Code | ~/.claude/skills/ (Windows: C:\Users\<you>\.claude\skills\) |
| ZCode | ~/.zcode/skills/ |
| OpenClaw | your active workspace's skills directory |
| Other Agent-Skills-compatible tools | that tool's skills directory |
2. Get your Account-SK — sign up / log in at the entrance convenient for you: mainland China open.sorftime.com (phone/WeChat, CNY) or international open-intl.sorftime.com (email/Google, PayPal/Stripe); both grant free trial credits → open the MCP 服务 page → subscribe → copy the Account-SK. One subscription covers all platforms; quota is metered per call.
3. Tell your AI: "Configure Sorftime for me, my key is " — it installs, verifies, and runs your first query. Manual alternative: python scripts/install.py --key <KEY> (Windows: use python, not python3).
Prefer the official one-click path? The MCP 服务 page generates a copy-paste instruction for your AI (single- or multi-platform), configuring native MCP servers named sorftime-<platform> — this skill then automatically serves as the knowledge layer for data queries (no extra key setup needed; batch scripts still want a one-time local key via install.py).
Skill family: this repo is the all-platform master skill; the official download slot oss.sorftime.com/download/skill/sorftime-mcp.zip is being updated to this build (it currently serves an older v3.x). Platform-specific editions (sorftime-amazon, sorftime-walmart, sorftime-shopee, sorftime-tiktok, sorftime-temu, sorftime-ali1688, sorftime-mercado, sorftime-social) are built from the same source — same Account-SK works everywhere.
- 🎯 v3.6.0 Instruction Priority — Obey the Seller: Precise instructions execute directly (no routing, no methodology cards, no lens imposition) — ask for "product_search yoga mat sorted by sales, data only" and you get exactly that, unfiltered and unembellished. Ambiguous/exploratory requests still get the guided flow. The Closed-Loop workflow is now an opt-in reference (runs only via
/goal//loopor explicit request); daily queries never enter it. - 🔇 Risk folding for pros:
pro/factory/brandprofiles collapse the risk table to a footnote + one-line 🔴hard hints (--show-risksto expand);newbie/growerkeep full warnings + safe-shortlist. Risk stays advisory — never blocks, never hides. - 🛡️ v3.5.0 Advisory Risk Model — The 4-tier risk system is now advisory, not blocking. Products are never hidden. Every result carries a risk badge (
🔴hard/🟡capital/🟠ops/⚠️trap/🟢safe) with a specific warning, surfaced AFTER the results. The seller — or an independent review agent — makes the final call. This kills the false-positive blocking other tools suffer from: non-slip yoga mats are no longer "apparel", alcohol wipes are no longer "chemicals", microwave-safe containers are no longer "large appliances". Matches the skill's own "no hard thresholds, full-set ranking" principle. - 🧐 Independent Shortlist Review — New
scripts/review_shortlist.pyruns a deterministic second-opinion (category-relevance, margin sanity, IP/trademark, review anomalies, seasonal windows, compliance red flags) emittingGO / CAUTION / NO-GO. SKILL.md §1.1.5 adds a mandatory post-selection review protocol: Layer 1 script + Layer 2 fresh-context independent sub-agent red-team.picker.py --jsonwires discovery → review seamlessly. - 🔀 mcp SDK 1.x/2.x dual-support — The local bridge now runs natively on both mcp 1.x and 2.x (version-branched Server API, per the official 2.0 migration guide), plus a self-heal guard for unsupported SDK majors. No more "MCP 2.0 not yet supported" errors on any device — the skill self-repairs on first start.
- 🧪 Agent-level verified — 3 independent agents exercised the skill end-to-end (beginner blue-ocean, pro competitor teardown, Walmart discovery), surfacing and fixing real doc-vs-code gaps (keyword-pollution false-GO, non-existent
--moderefs,product_reviewspagination). - Full changelog →
August 4, 2026 updates
- 🔄 Closed-Loop Product Selection Workflow — The flagship end-to-end pipeline. Covers product discovery → supply chain → financial analysis → risk assessment → 7-member independent AI Seller Review Panel → Go/No-Go deliverable → post-launch monitoring. Dual-path (HPI Product-First + Market-First with auto-routing). 6 business models, 7 platform adapters. One command:
/goal /sorftime-seller-agent Execute Closed-Loop Product Selection for yoga mats on Amazon US. Seller: beginner, $10000, private-label.Wiki · Case Study - 🤖 7-Member Independent Seller Review Panel — GO/CAUTION/NO-GO verdicts are no longer algorithmic thresholds. 7 independent AI sub-agents, each role-playing a different seller perspective (Peer, Mentor, Conservative, Opportunity, Platform Specialist, Financial Auditor), vote on every product with weighted scoring. Platform Specialist and Financial Auditor hold veto power. Panel may override algorithm. Validated: 20 rounds, 82.7% avg adoption.
- 💾 Data Persistence — Every phase auto-saves to
${SORFTIME_OUTPUT_DIR}. 9 files per run: raw API data, P&L calculations, panel votes, self-contained Markdown deliverable, copy-paste-ready monitoring command. Cross-platform. - 📊 product_traffic_terms field trap documented — The
exposure_positionfield ("Organic"/"Ad"/"Ad,Organic") is the correct way to assess organic traffic. The non-existentorganic_searched_percentagefield returns false zeros. Now hard-coded into the skill's gotchas.
August 3, 2026 updates
- 🧠 HPI 5D Signal Scoring workflow — Built-in skill workflow. Finds products with low ad dependency, low reviews, stable pricing, growing sales, and climbing BSR. 10 categories tested, 71 verified products.
- ⚡ 58 Loop & Goal templates — Wiki with ready-to-run commands. HPI verified shortlist, full-market blind scan, category cross-mining.
July 31, 2026 updates
- 🛡️ MCP 2.0 compatibility hardened, 📚 Wiki live, 1688 factory sourcing, Shopee cross-border, Amazon + Walmart audit.
注册/Registration: 大陆 open.sorftime.com（手机号/微信，人民币）· International open-intl.sorftime.com（email/Google, PayPal/Stripe）。新账号均有免费体验。
git clone https://github.com/DannylydST/sorftime-seller-agent.git
cd sorftime-seller-agent
python3 scripts/install.py"Find blue ocean products in yoga mats on Amazon US for a beginner"
"Analyze this Amazon competitor: ASIN B07H9PZDQW — traffic keywords, pricing, sales trend"
"Calculate my Amazon FBA profit: price $29.99, cost $8.50, weight 1.2lb"
"Pull Shopee MY phone case category Top 20 — who's selling, what prices, brand share"
"Compare this product's price between Walmart and Amazon"
| If you are a… | You can… |
|---|---|
| Amazon FBA seller | Product discovery, competitor reverse-ASIN, keyword research, profit calculator, listing audit, review mining |
| Walmart marketplace seller | Category scanning, keyword ranking, product trend tracking, cross-platform price gap analysis (Walmart vs Amazon) |
| Shopee / TikTok Shop / TEMU seller | Category research, market analysis, product trend monitoring, shop competitor intelligence |
| Dropshipper / arbitrage seller | Cross-platform price comparison, 1688 factory sourcing, multi-marketplace opportunity scanning |
| E-commerce agency / VA | Batch competitor analysis, keyword library building, automated monitoring for clients |
| Use Case | Example Prompt | |
|---|---|---|
| 🔍 | Amazon Product Research | "Find blue ocean kitchen products under $30 on Amazon US for a beginner seller" |
| 💎 | Hidden Profit Index ⭐ | "Scan the yoga mat category — surface products with high hidden profit potential that other tools miss" |
| 🎯 | Competitor Analysis (Reverse ASIN) | "Break down ASIN B07H9PZDQW — monthly sales, traffic keywords, pricing history, FBA fees" |
| 🔑 | Amazon Keyword Research | "What are the best long-tail keywords for yoga mats? Show search volume and competition" |
| 💰 | Amazon FBA Profit Calculator | "Calculate FBA profit: $29.99 selling price, $8.50 unit cost, 1.2lb weight" |
| 📊 | Market & Category Analysis | "Analyze the blender category on Amazon US — brand monopoly, pricing trends, market gaps" |
| 🏪 | Shopee Market Research | "Pull Shopee MY phone case category Top 20 — brands, pricing bands, shop types, sales distribution" |
| 🛒 | Walmart Product Research | "Find blue ocean products on Walmart US under $25 with low review counts" |
| 🎬 | TikTok Shop Intelligence | "Show me top-selling TikTok Shop US beauty products and their video authors" |
| 📈 | Automated Price & Sales Monitoring | "Watch ASIN B07H9PZDQW and alert me when price drops below $15 or sales spike 50%" |
| 🌏 | Cross-Platform Arbitrage | "Find products priced 30%+ higher on Walmart than Amazon US — low competition on Walmart side" |
Amazon sellers spend $29–$229/month on marketplace research tools. You log into dashboards, click through menus, export CSV files, then manually analyze in spreadsheets.
Worse — most tools only show you what you ask for. Filter by minimum reviews, filter by minimum price, sort by sales. A product with 13 reviews, 106 units/month, healthy margins, and zero ad competition gets filtered out before you ever see it. That's the product you should be looking at.
Sorftime Seller Agent replaces that workflow with conversation — and finds the products other tools hide from you.
Before: Log in → navigate → click → export → analyze manually (15–30 min)
After: "Find blue ocean yoga mat products on Amazon US" → results in 20 seconds
| Traditional Seller Tools | Sorftime Seller Agent | |
|---|---|---|
| Interface | GUI dashboards | AI conversation (natural language) |
| Platforms | Amazon only (typical) | 6 platforms — Amazon, Walmart, TikTok Shop, Shopee, TEMU, 1688 |
| Data depth | Basic product metrics | 160+ dimensions + 20 proprietary indices including Hidden Profit Index ⭐, Blue Ocean Finder, and 17+ methodology cards |
| AI integration | None — manual operation | MCP-native — any AI agent can query and analyze |
| Works with | Browser only | Claude Code · Codex · Cursor · OpenClaw · Hermes · Pi · any MCP agent |
| Automation | Manual workflows | Agent auto-execution — scheduled monitoring, batch analysis |
| Pricing | $29–$229/month | Free trial + usage-based paid tiers |
| Setup | 20+ minutes | Under 3 minutes |
| Open source | Closed source | MIT license — inspect, modify, contribute |
Not a Claude Code plugin. Not an API wrapper. A self-contained MCP bridge + intelligence layer that any MCP-compatible AI agent can load and use immediately.
| Agent / IDE | Setup |
|---|---|
| Claude Code | Auto-detected by mcporter, one-command install — under 3 min |
| Codex (OpenAI) | Import as MCP server — under 5 min |
| Cursor | MCP config auto-generated by install.py — under 3 min |
| OpenClaw | Native skill directory support — under 3 min |
| Hermes / Pi / any MCP agent | Standard tools/list → tools/call protocol — under 5 min |
What makes it agent-agnostic:
- MCP-first architecture — 97 tools, auto-synced with Sorftime server. Pure MCP protocol.
- Zero agent-specific code — No Claude-only prompts. No Cursor-only config.
- Auto-detection —
install.pydetects your environment and outputs the right config snippet. - Self-contained — Ships with its own Python venv, bridge, cache, and schema store. Python 3.10+ is the only dependency.
- Bilingual — English and Chinese input, auto-routed to the right analysis pipeline.
Hidden Profit Index (Sorftime exclusive ⭐)
You: "Show me products quietly making money on Amazon US — the ones other tools miss."
Agent: [potential_product — Hidden Profit Index scan] "Found a non-slip yoga mat: $29.99, 4.6 stars, only 13 reviews, 106 units/month. Near-zero ad competition. Price stable, category rank climbing. Every other tool filters this out because 'reviews < 50.' Hidden Profit Index: 13.44 — top of the category. Low barriers, healthy margins, real demand. The kind of product most sellers never see."
You: "Find blue ocean products in kitchen storage on Amazon US. Beginner seller, $5K budget."
Agent: [category scan + competition filter] "5 opportunities with low brand concentration, <500 reviews on top listings, healthy margins. #1: Under-sink organizers — 15K monthly searches, only 3 competitors with >100 reviews, estimated margin 35–42%."
You: "Reverse ASIN B07H9PZDQW — traffic keywords, pricing strategy, sales trend."
Agent: [ProductDetail + TrafficTerms + CompetitorKeywords] "Gaiam Yoga Block — 40K monthly sales, 4.7 stars, $18.99. Top traffic keywords: 'yoga block' (52K), 'yoga blocks 2 pack' (28K), 'cork yoga block' (15K). FBA fee: $12.82. Top competitor ASINs: B07XJ5YV8G, B09XMQ5J8W."
You: "Sourcing yoga mats at $8.50. Selling at $29.99, 1.2lb. Worth it on Amazon FBA?"
Agent: [FBA fee estimation + profit calculation] "Gross margin: 38%. Break-even: 9 units/day. Return rate sensitivity: profit stays positive up to 12% returns. Verdict: viable — proceed."
You: "Keyword strategy for a new yoga mat listing on Amazon US."
Agent: [KeywordDetail + KeywordExtends + KeywordSearchResults] "Top opportunities: 'extra thick yoga mat' (147K searches, low competition), 'non slip yoga mat' (98K), 'yoga mat with strap' (52K). Avoid 'yoga mat' (1.4M, saturated — Amazon Basics + Gaiam dominate)."
You: "Shopee Malaysia phone case market — Top 20 products, pricing bands, shop types."
Agent: [shopee_category_request] "20 products, RM2.38–RM102.37. Ugreen Flagship Store leads with 3.5K monthly sales at RM13.90. 13 brands, all regular shops (no flagship in Top 20). 9/20 products under RM5 — low-end dominates."
You: "Find products selling 30%+ higher on Walmart than Amazon US."
Agent: [Cross-platform gap scan] "3 products with significant gaps. 'Premium Yoga Block Set' — $34.99 Walmart vs $24.99 Amazon (40% premium). Only 3 Walmart sellers vs 12 on Amazon — low competition entry point."
| Platform | Data Available | Sites |
|---|---|---|
| Amazon | Products · Keywords · Categories · Reviews · Traffic · Trends · FBA Profit | 14 sites (US, GB, DE, FR, IT, ES, JP, CA, MX, AU, IN, AE, SA, BR) |
| Walmart | Products · Keywords · Categories · Traffic · Trends · Variations | US |
| TikTok Shop | Products · Categories · Authors · Videos · Trends | 8 sites (US, GB, ID, JP, MY, PH, TH, VN) |
| Shopee | Products · Keywords · Categories · Shops · Trends · Shop Intelligence | 8 sites (MY, PH, VN, TH, ID, SG, TW, BR) |
| TEMU | Products · Categories · Shops · Trends | US, EU |
| 1688 | Products · Variations · Similar items | CN |
Prerequisites: Python 3.10+ · Free Sorftime account (mainland) / open-intl.sorftime.com (international) · Any MCP-compatible AI agent
git clone https://github.com/DannylydST/sorftime-seller-agent.git
cd sorftime-seller-agent
python3 scripts/install.py # One-click: venv, deps, MCP config
python3 scripts/healthcheck.py # VerifyPlatforms: macOS · Linux · Windows (Python 3.10+)
Q: Is this official Sorftime? Yes. Built by @DannylydST at Sorftime Data Technology. Connects to Sorftime's official MCP API at open.sorftime.com.
Q: Is it really free? 新账号可免费体验 — 无需信用卡。高用量卖家/机构可通过人民币购买更高档位。
Q: Which AI agents can I use?
Any MCP-compatible agent: Claude Code, Codex (OpenAI), Cursor, OpenClaw, Hermes, Pi, and more. install.py auto-detects your environment.
Q: Where do I get my MCP Key? https://open.sorftime.com（大陆）或 https://open-intl.sorftime.com（国际）注册 → MCP 服务页 → 购买开通 → 复制 Account-SK。
Q: Can I use this without an AI?
Yes — all scripts work standalone: python3 scripts/picker.py --keyword "yoga mat". But AI-driven analysis with the 20 methodology cards unlocks the full value.
Q: What methodology cards are included?
Hidden Profit Index, Blue Ocean Finder, Competitor Deep-Dive, Keyword Strategy, Cross-Platform Price Gap, and 15 more. Full-ranking models — no hard thresholds that hide borderline opportunities. See references/methodology-cards/.
Q: Does this work for non-US Amazon marketplaces? Yes — all 14 Amazon sites: US, GB, DE, FR, IT, ES, JP, CA, MX, AU, IN, AE, SA, BR. Plus Walmart US, 8 Shopee regions, 8 TikTok regions, TEMU US/EU, and 1688 CN.
- Wiki — Platform guides, methodology cards, troubleshooting, glossary
- 📚 Product Selection Methodology (Multi-Platform) — The unified seller framework: shelf vs. interest-based e-commerce, resource screen → full-set ranking → validation → 3-scenario ROI → build → monitor & exit. Includes live Nov 2026 cases (iPhone 17 case CPC 2.2× in 19 months, Covers 10× seasonality, TikTok cleaning-brush creator matrices)
- ⚡ Loop & Goal Templates — 58 automation recipes: daily monitoring, weekly refresh, monthly Hidden Profit Index scans
- SKILL.md — Full skill reference: tools, parameters, gotchas
- CHANGELOG.md — Version history
- CONTRIBUTING.md — How to contribute
- SUPPORT.md — Getting help
This skill requires MCP SDK 1.x (not 2.x — breaking changes). The requirements.txt locks to mcp>=1.0.0,<2.0.0.
# If you hit "MCP 2.x is not yet supported" or AttributeError on startup:
python3 scripts/install.py --upgrade
# This force-reinstalls all dependencies at the correct versions.git pull # Pull latest skill updates
python3 scripts/install.py # Rebuild venv (if deps changed)
python3 scripts/healthcheck.py # Verify everything works| Symptom | Solution |
|---|---|
AttributeError: 'Server' object has no attribute... |
python3 scripts/install.py --upgrade |
MCP 2.x is not yet supported |
python3 scripts/install.py --upgrade |
ModuleNotFoundError: mcp |
Run python3 scripts/install.py |
| Connection test fails | Check MCP Key, network, firewall |
MIT © DannylydST · Sorftime Data Technology
Built for sellers. By sellers. On 6 platforms.