{"slug": "automating-weekly-product-research-with-claude-and-mcp", "title": "Automating Weekly Product Research With Claude and MCP", "summary": "A developer built an open-source MCP server called sorftime-seller-agent that lets Claude automate weekly product research for marketplace sellers. By connecting Claude to live data from Amazon, Walmart, TikTok Shop, and other platforms, the tool reduces a 45-minute manual workflow to a single 30-second query. The setup takes under five minutes and requires only a free API key.", "body_md": "Every Monday you open 4 tools, run the same queries, compile the same spreadsheet. You have been doing this for years. What if your AI did it for you in 30 seconds?\n\nThat is the promise of connecting Claude to live marketplace data through MCP -- the Model Context Protocol. Instead of logging into multiple dashboards, exporting CSV files, and stitching together a market picture by hand, you ask Claude one question and get a structured answer backed by real-time data. No switching tabs. No copy-paste. No pivot tables at 8 AM.\n\nThis article walks through how it works, what you can build with it, and how to set it up in under five minutes using the open-source `sorftime-seller-agent`\n\nMCP server.\n\nAsk any marketplace seller what Monday morning looks like and the pattern is nearly identical across categories, revenue levels, and platforms. The player changes -- Amazon, Walmart, TikTok Shop, Shopee -- but the choreography does not.\n\nTypical weekly research checklist:\n\nEach step is 5 to 15 minutes of clicking through filters, date ranges, and export buttons. The cognitive load is not the analysis -- the analysis takes 30 seconds once the data is in one place. The load is the *assembly*. Assembling data from fragmented tools is where the hours disappear.\n\nMCP lets Claude talk directly to external data sources as if they were built-in functions. Instead of you reading a dashboard and telling Claude what you saw, Claude reads the data itself and answers your question.\n\nHere is a concrete example. You type this prompt into Claude after connecting `sorftime-seller-agent`\n\n:\n\n\"Check my tracked ASINs on Amazon US. Which competitors dropped price by more than 5 percent this week? Which keywords moved more than 10 positions? Flag any product with margin below 15 percent after fee changes. Output a table sorted by urgency.\"\n\nClaude calls the relevant MCP tools -- fetching product details, keyword rankings, fee estimates -- combines the results, and returns a formatted table. Total time: roughly 30 seconds. The same analysis done manually: 45 minutes minimum, spread across three browser tabs and a spreadsheet.\n\nThe difference is not speed alone. It is that the bottleneck moves from data collection to decision-making. You stop being a human ETL pipeline and start being a strategist again.\n\n`sorftime-seller-agent`\n\nis an open-source MCP server that exposes marketplace intelligence as tools Claude can call. It covers over 130 endpoints across Amazon, Walmart, TikTok Shop, Shopee, Temu, and 1688.\n\nHere is what Claude can do once the server is connected:\n\nEach capability is a discrete tool in the MCP server. Claude decides which tools to call based on your question. You do not need to know the endpoint names or parameter formats -- Claude handles that automatically.\n\nThe setup is three steps and takes under five minutes. You need Python 3.10 or later and a Claude client that supports MCP (Claude Desktop, Claude Code, or any MCP-compatible client).\n\n```\ngit clone https://github.com/DannylydST/sorftime-seller-agent\ncd sorftime-seller-agent\npython3 scripts/install.py\n```\n\nThe install script sets up a Python virtual environment, installs dependencies, and generates a starter configuration file.\n\nAdd the server to your Claude client's MCP configuration. For Claude Desktop, edit `claude_desktop_config.json`\n\n:\n\n```\n{\n  \"mcpServers\": {\n    \"sorftime-seller-agent\": {\n      \"command\": \"python3\",\n      \"args\": [\"-m\", \"sorftime_seller_agent\"],\n      \"env\": {\n        \"SORFTIME_API_KEY\": \"your-api-key-here\"\n      }\n    }\n  }\n}\n```\n\nFor Claude Code, add it to `.claude/mcp.json`\n\nin your project or home directory with the same structure.\n\nSign up at [open-intl.sorftime.com](https://open-intl.sorftime.com) to get a free API key. The free tier includes enough quota to run weekly research workflows on a standard product portfolio.\n\nRestart your Claude client. You should see a hammer icon or tool indicator confirming the MCP server is connected. Ask Claude: \"List the tools you have access to.\" If `sorftime-seller-agent`\n\ntools appear, you are ready.\n\nOnce connected, here is a prompt you can use as a starting template. It covers the four core areas of a weekly review in a single query.\n\n\"Run my weekly product health check. For all ASINs I am tracking:\n\n- Compare current BSR and price against last week. Flag changes over 10 percent.\n- Pull keyword rankings for my top 5 target terms per ASIN. Note any position shifts.\n- Check for new competitors that entered the top 20 BSR in each ASIN's primary category.\n- Recalculate profit margins using current FBA fee estimates and my standard COGS. Highlight anything below 18 percent.\nFormat the output as a markdown table with one row per ASIN and a one-line verdict for each.\"\n\nClaude executes this by calling the relevant tools in sequence, waiting for each result before making the next call, then synthesizing everything into a single structured table.\n\nThe first time you run it, review the output and refine the prompt -- add your specific margin thresholds, your competitor watchlist, the keywords that matter to your category. By the third week, the prompt is dialed in and Monday morning research takes the time it takes to type \"run my weekly product health check.\"\n\nThere is a deeper reason this approach is gaining traction among marketplace sellers. Traditional SaaS dashboards are designed for *human* data consumption -- charts, tables, filters, export buttons. They optimize for a person looking at a screen and making decisions click by click.\n\nMCP flips the model. Data flows directly into the reasoning engine. Claude does not need a chart. It needs structured numbers, and it is better at spotting patterns across multiple data dimensions than a human scanning a dashboard.\n\nThis also means your research scales without adding headcount. Tracking 10 ASINs or 100 ASINs is the same prompt. Claude handles the fan-out, calling tools for each ASIN and aggregating results. What changes is execution time, not your effort.\n\nA few honest boundaries. The MCP server provides data access and tool definitions. It does not make business decisions for you. It will not tell you which product to launch. It will not guarantee a keyword ranks or a margin holds. Market conditions shift, fees change, and data always has a freshness window.\n\nThe value is in removing the repetitive assembly work so you can spend your attention on the decisions that require judgment. Think of it as moving from \"person who collects data\" to \"person who interprets data.\"\n\n```\ngit clone https://github.com/DannylydST/sorftime-seller-agent && cd sorftime-seller-agent && python3 scripts/install.py\n```\n\nFree API keys are available at [open-intl.sorftime.com](https://open-intl.sorftime.com). The source code is open -- read it, fork it, extend the tools to match your workflow. If you build something useful, open a pull request.\n\nOne prompt. Thirty seconds. No more Monday spreadsheet assembly.", "url": "https://wpnews.pro/news/automating-weekly-product-research-with-claude-and-mcp", "canonical_source": "https://dev.to/tomi431/automating-weekly-product-research-with-claude-and-mcp-2plo", "published_at": "2026-07-28 08:01:41+00:00", "updated_at": "2026-07-28 08:04:25.899595+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "developer-tools", "ai-products"], "entities": ["Claude", "MCP", "sorftime-seller-agent", "Amazon", "Walmart", "TikTok Shop", "Shopee", "Temu"], "alternates": {"html": "https://wpnews.pro/news/automating-weekly-product-research-with-claude-and-mcp", "markdown": "https://wpnews.pro/news/automating-weekly-product-research-with-claude-and-mcp.md", "text": "https://wpnews.pro/news/automating-weekly-product-research-with-claude-and-mcp.txt", "jsonld": "https://wpnews.pro/news/automating-weekly-product-research-with-claude-and-mcp.jsonld"}}