Topview MCP does not make a store sell by itself, and its generated actors still look generated. What it does well, on the creator's own test, is force a product decision before you spend on ads, which is where he found the real value.
What is Topview MCP, and what does it actually run inside Claude Code?
Topview MCP is a hosted Model Context Protocol server that adds market research, campaign planning, image, storyboard and video generation to an agent such as Claude Code. The creator, Mano Deyvin, installed it in a project's .mcp.json file, authenticated through mcp authenticate, and then called its tools directly from chat.
Model Context Protocol is the open connector format that lets a chat agent call outside services as tools. Topview's own setup page documents the configuration for Claude Code and other MCP-capable clients, plus an npx skill installer the creator also tried. Authentication happens through a browser consent screen, and the terminal confirms success before any tool call works.
Two packages exist here: the MCP server that exposes tools, and the skills that package instructions for recurring jobs. Neither one is an ecommerce platform. The creator kept his store elsewhere, and Topview never touched hosting, payments or cart.
The creator also used Topview Canvas, a visual workspace inside Topview's own interface that pulls the generated assets into one board. That part lives outside Claude Code, and he opened it through a browser link the agent produced.
Which Topview MCP tools did the creator actually use?
The creator's session used a research tool, a campaign-planning output, an image generator, a storyboard generator, a video generator and a canvas exporter. He never showed a complete tool list in the video, so treat this as the set that appeared rather than the full catalog.
The workflow ran in a fixed order, and each step fed the next one. He asked for research first, saved it, then built the campaign on the saved findings, then generated mockups from the approved design, then condensed the storyboard, then turned frames into video clips, then exported everything to the canvas.
- Research: output named relevant channels, their promoted products, search-term patterns and absolute purchase counts.
- Asset saving: outputs were written to Markdown files so later steps could reuse them.
- Image generation: three hat mockups from one logo upload, reportedly routed across two underlying image models.
- Storyboard: a plan condensed from six scenes to three.
- Video generation: clips rendered at 1080p and 720p.
- Canvas export: one link collecting approved mockup, storyboard and finished clips.
The model routing in the video matters for expectations. The creator says one mockup came from Nano Banana Pro and two from a GPT image model, but that is his reading of the agent's own logging, not a documented routing rule he controlled.
What the market research found about tech merch
The research output told the creator that printed t-shirts search heavily but convert weakly, while embroidered hats search far less and sell more. He then changed his launch lineup from t-shirts to hats, hoodies and a large mousepad.
The specific numbers came from the agent's research tool, describing Amazon search and purchase data. The creator did not independently audit the dataset, so treat the figures as tool-reported. The tool put printed t-shirts at 83,424 monthly searches with almost no purchases, and said embroidered hats see roughly eight times fewer searches but 5.5 times more purchases in absolute count.
The tool also said it had tested 26 search terms while examining whether channel-branded merch sells through search at all, and reported that specific creator-merch terms were not a viable search channel. The consequence for the creator was practical: a storefront link under videos, not a keyword-led store.
One finding landed because of his existing product. He owns a hat carrying his channel's 33 logo, and the tool flagged that heavy embroidery loses fine detail such as the eye inside the symbol. He accepted that change to a simplified solid version.
From research to 5-hook campaign plan and storyboard
The campaign step produced five opening hooks for an embroidered hat, each aimed at a different angle: wardrobe comedy, wash durability, quality close-up and workplace humor. The creator's saved plan names the chosen product, the hook, the video concept and the storyboard direction.
The hooks read the way a decent social copywriter writes them. One plays on developers owning many conference shirts and no good hat. Another closes on washability as the proof point. A third frames the hat as merchandise that survives the laundry machine. The creator reacted most to the washability line, and he noted the research had read his uploaded logo image before proposing the simplified embroidery.
The storyboard then compressed into three scenes: a hook, a body and a call to action. The body scene shows someone opening a wardrobe full of tech conference shirts from companies including AWS, Microsoft and Google, then finding the hat. The agent generated one version, judged it too blurry to read the shirts, and generated a second pass.
How good were the generated images and videos?
Generated images held up better than generated video. The creator liked the first hat mockup most, and the second storyboard pass fixed the readability failure he had seen. Video clips rendered at 1080p and 720p, and he judged them passable when no human face fills the frame.
His verdict on faces was blunt: the actor shots still look AI. The wardrobe scene and the finger-across-the-logo close-up stayed convincing because they avoid a full face. He also pointed out that the model produced a 720p version even though he could not find a tool to cancel the 1080p render, so he paid the queue time in both.
The 1080p and 720p clips of the same call-to-action scene came out close but not identical, with slightly different camera movement. That detail matters if you plan to intercut versions, because they are separate generations rather than two encodes of one render.
He called the whole output a bazooka for an ant if you have no store, no product and nothing to sell. For someone already selling a service or running a store, the same session replaces work he was quoted R$15,000 for at the Forum E-commerce Brasil.
What Topview MCP costs and where it stops
Pricing, quota and credit balances were not shown in the video, and Topview's commercial terms can change, so check the vendor's current plan page before budgeting. The creator offers affiliate credits in his description, which is a commercial connection worth knowing when you weigh his review.
The tool's limits are more interesting than its price. It wrote Markdown into his project because the agent chose to, not because Topview enforces file output. Asset storage, resolution choices and rerun behavior come from the underlying models, and those decisions were partly opaque to him during the session.
Coding-adjacent content fits this workflow, but the tools describe ecommerce and campaign work rather than software delivery. The creator's own framing is that selling is not just code, payment and deploy, and this MCP covers the part that used to sit with agencies.
For a developer learning in public, the companion programs from the same creator ecosystem are separate paid products. He points viewers toward Gustavo Dev Doido, the Crazystack Typescript track and the Bootcamp do Dev Doido as the programming side of his content. Those links are not required to use Topview.
Topview MCP vs an agency package
An agency package quoted at R$15,000 bundled research, creative angle, ad script and finished video. Topview MCP produced the same four categories of output in one chat session, with the creator's own judgment replacing an account manager.
The comparison is not a clean equal. An agency crew delivers footage of real people, and the creator said plainly that the generated actor scenes look artificial. He also said he could shoot the simple storyboard himself in a grey shirt and get better ad footage, which is the strongest argument for using the storyboard as a shooting script rather than a finished ad.
| Dimension | Topview MCP | Agency package |
|---|---|---|
| Quoted cost in the video | Not disclosed | R$15,000 |
| Market research | Tool-reported, minutes | Human analyst, days |
| Creative and hooks | Generated in chat | Human strategist |
| Finished video | Generated clips, AI-looking faces | Filmed actors |
| Editing control | Canvas workspace | Producer and editor |
| Best use | Fast test before spending | Full campaign delivery |
If you want a decision from this test: use the MCP research and hooks, then film the cheap scenes yourself. That combination is what he described, and it keeps the parts that passed.
Questions people ask about Topview MCP
- Does Topview MCP work with Claude Code only? No. Topview documents setup for several MCP-capable clients, and the creator shows a Claude Code configuration plus an
npxskill install path. Any client that supports MCP servers with browser authentication should connect. - Did the creator verify the market research numbers? No. He described the search and purchase figures as the tool's output and said he would research further, so the 83,424 searches and the 5.5-times purchase contrast remain tool-reported.
- Can Topview MCP generate a complete ad without any filming? In the test it produced mockups, a storyboard and short clips at 1080p and 720p. The creator still judged the human scenes as visibly generated and said he could film them himself.
- Is Topview affiliated with the brands whose shirts appear in the storyboard? Nothing in the video suggests any relationship. Logos such as AWS, Microsoft and Google appear as wardrobe props in a generated scene.
- What should I try first if I sell something? Start with the research step on your actual product line, save the output to a file, and only then ask for a campaign plan. The creator's sequence worked because each step reused saved findings.
Turn the video you already recorded into the article
The research part of this test mattered more than the rendered clips, and the same pattern shows up in every good YouTube video: a real decision, a real failure, and the reasoning between them. That knowledge usually stays locked in the audio track.
If you have already explained a product choice, a tool test or a lesson on camera, Skala Blog turns that video into a written article. Paste a YouTube URL, let it transcribe the video, and generate a structured article you can edit and publish.