# Can Claude make images? How to generate images, video and product shoots in Claude Claude can't make photos on its own. Connect Sloosh over MCP and it can make images, video and full product shoots, with a quote before every paid run. 21 Oct 2026 · 8 min read · Guides · By Sloosh Web page: https://sloosh.ai/blog/generate-images-in-claude To generate images in Claude, connect an image tool to it over MCP. Add Sloosh at `https://mcp.sloosh.ai/mcp`, sign in, and set a daily credit limit. Then ask in plain words. Claude picks one of 24 image models or 22 video models, or a ready-made Space, shows you a quote, and hands back the finished file. ## Key facts | | Sloosh MCP | | --- | --- | | Server address | https://mcp.sloosh.ai/mcp (same for every app) | | Works in | Claude (web, desktop, Claude Code), Codex, Cursor, VS Code, ChatGPT Business and Enterprise/Edu, and other tools that connect to remote MCP servers over HTTP | | Plans | Every plan. Prices at sloosh.ai/pricing | | Models | 24 image models and 22 video models. Defaults: Nano Banana 2.1 (image), Seedance 2.5 (video) | | Spending | Nothing paid runs until you set a daily limit. It resets at midnight UTC and covers every connected app | | Quotes | Every paid run is quoted first. A quote is good for 5 minutes | | Uploads | Upload a file, or import an https link, up to 64 MiB | | Where files land | Your assistant gets the result and its link; the file is saved in your Sloosh Assets | ## Can Claude make images on its own? Claude writes, reasons and codes. It can build a diagram, a chart or an SVG right in the chat. Anthropic's own help page says Claude [doesn't generate photos or illustrations](https://support.claude.com/en/articles/9002504-can-claude-produce-images) the way image-generation tools do. For a photo of your product on a marble counter, it needs a tool. That tool plugs in through MCP, the open standard that lets apps like Claude and ChatGPT call outside tools. Sloosh is an MCP server. Once it's connected, "make me a product shot" becomes a real file. ChatGPT can already make images. What Sloosh adds there is video, other image models, and Spaces. ## What can your assistant make once Sloosh is connected? Three sizes of job. 1. **One image.** A packshot, a banner, a texture for a mockup. 2. **One clip.** A video of up to 30 seconds from Seedance 2.5, Wan 3 or Kling 4.0, or a shorter one from MiniMax and others. 3. **A whole Space.** A Space is a finished workflow someone already built and tested. You fill in the inputs and it runs every step. The third is the one most people miss. Say you sell a ceramic mug. You have three phone photos, shot on a kitchen counter next to a kettle, and one photo of a cosy autumn kitchen you'd like it to live in. Launch is Friday. You type one message: **One message** ```text Here are three photos of my mug. Clean them up on white for Amazon. Then put the cleanest one in this autumn kitchen photo. Then make a 9:16 selfie video ad with the person in this photo holding it. ``` Your assistant finds [White Background Product Studio](https://sloosh.ai/spaces/white-background-product-studio) and runs it. The clutter goes. The mug sits on an even, solid white background. It passes the clean shot and your kitchen photo to [Product in Scene + Relight](https://sloosh.ai/spaces/product-in-scene-relight), which places the mug on a surface in that scene and relights it to match, so it looks like it was shot there. Then it runs [UGC Selfie Video Ad](https://sloosh.ai/spaces/ugc-selfie-video-ad): a 4 to 15 second phone-style clip with a hook, the mug up close, an honest reaction, your call to action and a voiceover in the language you pick. Three Spaces. One message. Each run gets its own quote before a credit moves. Use a photo of a person who agreed to be in your ad. That part is on you. ## How do you connect Sloosh to Claude, ChatGPT, Cursor, Codex or VS Code? Every app uses the same address and the same browser sign-in. You never paste a password or an API key. The short version is below; [Sloosh MCP for Claude, ChatGPT and Cursor](https://sloosh.ai/blog/sloosh-mcp-claude-chatgpt-cursor) and the [MCP docs](https://sloosh.ai/docs/mcp) have every detail. ### Claude (web and desktop) 1. Download the Sloosh plugin from [sloosh.ai/mcp](https://sloosh.ai/mcp). 2. In Claude, open **Customize**, then **Plugins**. Choose **Add**, then **Upload plugin**, and pick the file. 3. Open the plugin's **Connectors** tab. Choose **Add** if Sloosh isn't there, then **Connect**, and sign in. 4. In the connector's settings, set read-only tools to always allow. Keep the rest on ask. No plugin? Use **Customize**, **Connectors**, **Add custom connector**, name it `Sloosh` and paste the address. Custom connectors depend on your Claude plan ([Anthropic's guide](https://support.claude.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp)). On Team or Enterprise, an Owner adds it first. ### Claude Code **Terminal** ```text claude mcp add --transport http sloosh https://mcp.sloosh.ai/mcp ``` Then run `/mcp` and sign in. ### ChatGPT On ChatGPT web, open **Plugins**, choose the plus button, then **Add custom MCP server**. Name it `Sloosh`, paste `https://mcp.sloosh.ai/mcp`, set authentication to OAuth, accept the risk warning and choose **Create as a plugin** ([OpenAI's guide](https://help.openai.com/en/articles/12584461-connecting-custom-mcp-servers-in-chatgpt)). OpenAI gives full MCP, which Sloosh needs to generate and run Spaces, to ChatGPT Business and Enterprise/Edu workspaces, where an admin or owner adds the server. On Pro, ChatGPT allows read and fetch tools only. On any ChatGPT plan, you can use Sloosh through Codex instead (below). ### Cursor Add this to `.cursor/mcp.json` in your project, or to your global MCP file, then connect and sign in: **.cursor/mcp.json** ```text { "mcpServers": { "sloosh": { "url": "https://mcp.sloosh.ai/mcp" } } } ``` ### Codex **Terminal** ```text codex mcp add sloosh --url https://mcp.sloosh.ai/mcp codex mcp login sloosh --scopes spacelab:read,spacelab:author,spacelab:execute,user:org:read,offline_access ``` The second command opens your browser to sign in. Codex comes with every ChatGPT plan, and the Codex IDE extension uses the same server setup ([OpenAI's Codex MCP docs](https://learn.chatgpt.com/docs/extend/mcp)). ### VS Code Add this to `.vscode/mcp.json`, run **MCP: List Servers**, start `sloosh` and sign in: **.vscode/mcp.json** ```text { "servers": { "sloosh": { "type": "http", "url": "https://mcp.sloosh.ai/mcp" } } } ``` ### Last step, every app Open [MCP settings](https://sloosh.ai/mcp#agent-settings) and set a daily limit. Until you do, your assistant can browse and draft but can't spend. ## How does the daily spend limit work? You set one number per workspace. Every connected app shares it. The agent can't raise it, and reconnecting doesn't reset it. It resets at midnight UTC. A second setting, **Ask before spending more than**, decides when it checks with you. At 0, every paid run needs your yes. Set it to a small number and cheap runs go through on their own. Retries reuse a run key, so a dropped connection doesn't charge you twice. Cancel a run and finished outputs stay charged; unused credits come back when the run ends. ## Prompts to copy Paste these as written. Swap in your product. **1. The first test** ```text Make a studio photo of my water bottle on a clean white background. ``` **2. The full shoot** ```text Here are 3 photos of my product and a photo of a cosy autumn kitchen. Run White Background Product Studio on the product photos. Then use the cleanest result and the kitchen photo in Product in Scene + Relight. Quote each step before you run it. ``` **3. The UGC ad** ```text Run UGC Selfie Video Ad with these product photos and this creator photo. 9:16, 12 seconds, English voiceover. Hook: "I didn't think a mug could fix my mornings." End on "Link in bio." ``` **4. Ad variants** ```text Create three 9:16 ad variants for my summer sale on this sneaker. Same shoe, three different backgrounds, room at the top for a headline. ``` **5. The clip** ```text Turn my latest product image into an 8-second Seedance 2.5 clip. The camera pushes in slowly until the label fills the frame. Soft room tone, no music. Quote it first. ``` **6. The shopping trip** ```text Find a Space that makes product ad grids and tell me what a run costs. Don't run it yet. ``` **7. In Cursor, for a whole catalogue** ```text Read product names from data/products.csv. For each row, make a 1:1 hero image on white with Nano Banana 2.1. Quote the full batch first, then add each asset link to a new column. ``` **8. The check-in** ```text Which workspace are you connected to, and how much of today's limit is left? ``` That last one costs nothing. Run it first. For video prompting in depth, see the [Seedance 2.5 prompts guide](https://sloosh.ai/blog/seedance-2-5-prompts). ## Studio, Space or MCP: which should you use? | You want to | Best route | Why | | --- | --- | --- | | Try 20 looks and pick by eye | Studio in the browser | Grids, side by side, fast | | Run a proven job with your inputs | A Space | Steps already tested | | Stay in the app where you work | MCP | One message, the result comes back to you | | Chain several Spaces | MCP | The assistant passes outputs along | | Batch from a spreadsheet or repo | MCP in Cursor, Codex or Claude Code | It reads your files | | Build your own repeatable flow | Spacelab | You see every node | ## When is an MCP the wrong route? Sometimes the browser wins. Here's when. **You're exploring by eye.** Chat shows a handful of results at a time. If you want forty versions on one screen, open Studio. **You need pixel surgery.** Removing a stray hair or matching a Pantone exactly is retouching work. Use a photo editor. **You need a picture in ten seconds.** If ChatGPT's own image tool gives you what you need, use it. Sloosh earns its place when you need video, a specific model, or a multi-step Space. **Your team signs off on every spend.** Set **Ask before spending more than** to 0. Every run waits for a yes. Slower, but nobody gets surprised. **The chat times out.** Long video runs can outlast a chat turn. A timeout doesn't cancel anything. The file still lands in your Sloosh Assets. ## One message, one shoot Three phone photos on a kitchen counter. One message. White background, the mug in your autumn kitchen and a selfie ad by the time your tea goes cold. Try it with 20 free credits at [sloosh.ai](https://sloosh.ai). ## FAQ **Can Claude generate images without an MCP server?** Claude can build diagrams, charts and SVG in the conversation, but it doesn't make photos or illustrations the way image models do. Connecting Sloosh over MCP gives it 24 image models and 22 video models. Setup takes one address and a browser sign-in. **Which Sloosh plans include MCP?** MCP works on every plan, and your 20 starter credits work with it. Your first Sloosh workspace gets them free, and you can also use them to try Studio and Spaces in the browser. Current prices are at sloosh.ai/pricing. **Can my assistant overspend?** Only up to the daily limit you set, and it can't raise that limit. Every paid run starts with a quote that is good for 5 minutes. Set Ask before spending more than to 0 if you want to approve each run. **Where do generated files go?** Your assistant gets each result and its link, and the file is saved in your Sloosh Assets. Prompts pass through your AI app to the models you run.