Sorftime Seller Agent — Free AI-Powered Amazon Product Research & Marketplace Intelligence

<p align="center"> <img src="assets/hero-extension.png" alt="Sorftime — AI-powered Amazon product research, competitor analysis, keyword research, and marketplace intelligence for 6 platforms" width="100%"> </p> <p align="center"><em>9 years of seller data · 40+ global marketplaces · 160+ data dimensions · Backed by Sorftime's AI data supply chain</em></p> <p align="center"> <img src="assets/platform-logos.png" alt="Amazon · Walmart · TikTok Shop · Shopee · TEMU · 1688" width="100%"> </p>

AI-powered product research, competitor analysis, and keyword intelligence for Amazon, Walmart, TikTok Shop, Shopee, TEMU, and 1688 sellers.

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.

<p align="center"> <img src="assets/company-intro.png" alt="Sorftime — The most comprehensive overseas e-commerce data company globally. 97% market segment coverage across global platforms. Real-time consumer trend insights from social media. Core philosophy: marketplace data reflects transactions that already happened, social media reveals demand that is brewing." width="100%"> </p>

---

🆕 What's New (August 10, 2026)

• 🎯 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//loop or explicit request); daily queries never enter it.
• 🔇 Risk folding for pros: pro/factory/brand profiles collapse the risk table to a footnote + one-line 🔴hard hints (--show-risks to expand); newbie/grower keep 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.py runs a deterministic second-opinion (category-relevance, margin sanity, IP/trademark, review anomalies, seasonal windows, compliance red flags) emitting GO / 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 --json wires 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 --mode refs, product_reviews pagination).
• Full changelog →

<details> <summary>August 4, 2026 updates</summary>

• 🔄 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_position field ("Organic"/"Ad"/"Ad,Organic") is the correct way to assess organic traffic. The non-existent organic_searched_percentage field returns false zeros. Now hard-coded into the skill's gotchas.

<details> <summary>August 3, 2026 updates</summary>

• 🧠 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.

</details> <details> <summary>July 31, 2026 updates</summary>

• 🛡️ MCP 2.0 compatibility hardened, 📚 Wiki live, 1688 factory sourcing, Shopee cross-border, Amazon + Walmart audit.

</details>

---

⚡ Quick Start

1. Create a free account

open-intl.sorftime.com — sign up with Google, free trial credits included. PayPal for additional credits.

2. Install

git clone https://github.com/DannylydST/sorftime-seller-agent.git
cd sorftime-seller-agent
python3 scripts/install.py

3. Ask your AI

"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"

---

👤 Who Is This For

| 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 |

---

🧠 What You Can Do

| | 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" |

---

🤔 Why This Exists

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

---

🆚 How It Compares

| | 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 |

---

🤖 Agent-Friendly by Design

<p align="center"> <img src="assets/agent-friendly-demo.gif" alt="Same skill running in Claude Code, Codex, Cursor, and OpenClaw — one skill, four AI agents, six marketplaces" width="100%"> </p>

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 | |-----