Render your first preview

You will turn five raw Tidebook screenshots into one styled App Store preview on your own computer. The path is deliberately narrow: Local MCP, local PNG paths, one preview, one revision.

Turn these five Tidebook screenshots into an App Store preview. Keep them in filename order, use the supplied headlines, and show me the preview before making store-ready files.

#Before you start

You need:

• Local MCP connected through Your computer · Local MCP.
• Five raw PNG screenshots: 01_today.png, 02_forecast.png, 03_spots.png, 04_log.png and 05_alerts.png.
• The headline for each panel, in the same order as the files.

Use your own raw app captures. The Tidebook filenames and headlines keep this example concrete; replace the folder, files and words with yours. A screenshot that already contains a device frame or marketing caption will produce a phone inside a phone.

No ShotOps account is required for this preview. Local rendering reads the PNGs from your disk and uploads nothing.

Before continuing, ask “check my ShotOps account status.” A result that names the connection as Local confirms the client can reach the server. Follow the connection remedy in Troubleshooting if the tool is missing or does not run.

#Ask for a preview

Give the agent the absolute folder path and the five headlines. For Tidebook:

Render the PNGs in /Users/me/tidebook/screens as an App Store preview. Keep filename order. Use these headlines: “Know the tide before you go”, “Seven days of swell”, “Every spot you surf”, “Your own session log”, and “Wake up to the right swell”.

The agent calls render_strip with local { "path": "…" } screenshot entries and preview: true. You do not need to construct the arguments yourself.

That request names an order and headlines but nothing about colour or composition, so the agent may open with two styling questions — a palette and a layout — before it renders anything. Answer them, or say “surprise me” and it chooses. Naming a palette and a layout in the request itself skips the questions, and so does working on a project that already has a saved Look.

#Check the result

Success is visible: five panels come back in filename order, each with one Tidebook screenshot, one device treatment and the matching headline.

Before revising, check three things:

• The longest headline is readable and does not sit on the device.
• Every screenshot appears in the intended panel.
• The response does not report a layout finding or a missing source.

If a layout block appears, use its explanation and corrected Look where offered. If a source is missing, fix the path or filename and re-run the preview. Troubleshooting maps each observed condition to its remedy.

#Make one revision

The last Tidebook headline is the longest. Ask for one composition change in words:

Give the headlines more room by using the bleed layout, then render another preview.

The agent reuses the same screenshots and captions with the new layout. Compare the two previews and keep the one that communicates the screens more clearly. describe_look is where the full styling catalog lives — the agent reads it to ask those two questions, and you can ask to see the choices yourself when a word is not enough. The catalog itself stays out of this first path.

#What success looks like

You now have an approved five-panel preview and can state which layout to keep. The preview is a decision aid, not a store artifact: do not upload it to App Store Connect or pass it to fastlane.

This tutorial changed no saved project and created no durable hosted asset. The design exists only in the request and the returned preview images.

#Continue to store-ready output

• Work on a saved project preserves the screenshots, captions and Look so the strip survives the conversation.
• Plans and output explains what store-ready production requires and what Hosted rendering spends.
• Ship every locale applies the approved composition to each locale and produces the fastlane folder structure.
• Run it in CI adds deterministic inputs, durable Hosted recovery and safe cleanup.

This path is Local MCP throughout. If you are in ChatGPT, in CI, or working from screenshots that are not on this machine, ShotOps Cloud · Hosted MCP carries the same first preview end to end, without a shell or a local file path.