# Playground AI MCP > Playground AI lets your agent generate, modify, and upscale high-resolution images using only natural language instructions. Stop opening separate art programs; instead, tell your LLM to create assets—whether you need a simple placeholder or complex architectural rendering—and get the result instantly within your workflow. It handles everything from background removal to expanding image borders with precision. ## Overview - **Category:** ai-frontier - **Price:** Free - **Tags:** generative-art, image-generation, inpainting, outpainting, ai-models, creative-workflow ## Description This MCP connects your AI client directly to Playground AI's powerful compute clusters. Instead of manually dragging images into an external editor, you just instruct your agent what visual asset you need. You can ask it to generate entirely new graphics based on a text prompt, or take an existing sketch and refine it using specific guidance types like depth maps. If you need to fix a photo by filling in missing sections, the MCP handles that surgically. It's also perfect for scaling small icons up to 4x resolution without losing detail. When your AI agent needs this kind of visual power, connecting through Vinkius lets you access it from any compatible client, keeping your entire creative pipeline inside your existing code environment. ## Tools ### generate_image Creates new images from scratch using a text prompt, allowing you to set specific dimensions and select multiple models. ### transform_image Alters an existing image's style or content based on a prompt, controlling the degree of change via a strength setting. ### inpaint_image Fills in specific missing areas within an image using a provided mask and descriptive text. ### upscale_image Increases the resolution of an existing picture, enhancing both detail and overall clarity by a factor of 2 or 4. ### remove_background Separates a subject from its surroundings and returns a clean asset with a transparent background. ### outpaint_image Extends an image beyond its current borders, generating new content that matches the specified direction (up, down, left, right). ### generate_with_controlnet Creates images while forcing structural guidance using reference types like depth maps or edge detection. ### list_models Returns a list of all available AI models, detailing their capabilities and names for use in your prompts. ### get_generation Retrieves the full details, prompt, and metadata for a specific image generation ID. ### list_generations Lists records of your recent visual asset creations, including IDs and timestamps. ## Prompt Examples **Prompt:** ``` Generate a 1024x1024 image of a cyberpunk coffee cup in neon lighting. ``` **Response:** ``` Payload dispatched holding spatial constraints to 1024x1024 pixels. Using default Model 'Playground v3'. The system returns generation block `ID: 9x8A-CUP`. The finalized render URL is ready: [View Graphic]. Want to apply an iterative 2x Upscale on it now? ``` **Prompt:** ``` Upscale this image to 4x its size `https://example.com/small_icon.png`. ``` **Response:** ``` Upscaling protocol engaged. Submitted target to `upscale_image` invoking factor constraint '4'. The engine interpolated logical pixels correctly matching standard textures natively. Result pointer: [High-Res URL]. ``` **Prompt:** ``` Remove the background from the image at `https://example.com/person.jpg`. ``` **Response:** ``` Background segmentation dispatched via `remove_background`. The deep learning layer masked the subject heavily and omitted out-of-bounds layer fragments. I've received the pristine transparent PNG rendering back. The new asset is currently available at `[Extracted PNG URL]`. Would you like me to map it in a new CSS class? ``` ## Capabilities ### Generate images from text prompts Create new assets instantly by giving a simple description and specifying size constraints. ### Adjust existing visuals with guidance Change an image's style or content using a prompt while controlling how much the original image should be retained. ### Fill in missing parts of an image Use a mask to tell the AI exactly which area needs content, generating realistic fills over defined regions. ### Enlarge and expand images Scale blurry pictures up to 4x for high detail, or instruct the model to generate new background context outside the original frame. ### Isolate subjects from backgrounds Automatically remove a subject's surroundings and return a clean asset with a transparent background. ## Use Cases ### Building a Product Mockup Gallery A web developer needs to populate a gallery of placeholder product images. Instead of manually generating multiple assets, they prompt their agent: 'Generate six 1024x1024 images using `generate_image` depicting different views of the widget.' The MCP returns all necessary assets immediately. ### Marketing Campaign Asset Refresh A creative director has a product photo but needs to change the environment. They use `outpaint_image` on the original picture, prompting the AI to extend the background from an indoor setting into a sunny outdoor market scene. ### Cleaning Up Source Material A concept artist receives a messy sketch and needs to clean up the foreground character. They use `remove_background` first, then `inpaint_image` with an explicit mask over clothing details, ensuring structural consistency. ### Improving Low-Quality Inputs A team member has a low-resolution logo graphic that looks pixelated on the website. They pass the image URL to `upscale_image`, boosting it 4x and making it suitable for high-DPI screens. ## Benefits - It eliminates the need to manually open a separate art application. You simply instruct your agent, and it executes complex visual tasks like `generate_image` right in your workflow. - Achieve professional isolation of subjects instantly using the `remove_background` tool. This gives you perfectly clean assets without having to edit them in another program. - You can dramatically expand a photo's context using `outpaint_image`. If your scene is too narrow, tell the agent to generate the missing environment around the edges. - For maximum control, use `generate_with_controlnet`. This lets you guide the AI not just with text, but with structural inputs like depth or pose maps for reliable results. - Upscaling blurry or low-res assets is simple. Running `upscale_image` boosts quality by 2x or 4x, ensuring your final graphic looks crisp everywhere. ## How It Works The bottom line is that you treat complex visual editing like writing another function call; it’s just a prompt away. 1. Append the Playground AI integration to your MCP setup and provide your API key. 2. Instruct your agent using plain language (e.g., 'Generate a 1024x1024 image of...'). 3. The system executes the request, and you receive the final rendered asset URL back in your code output. ## Frequently Asked Questions **How do I use Playground AI MCP for image generation?** You initiate creation by using the `generate_image` tool. You simply pass a descriptive text prompt and specify the required dimensions (width and height) to get started. **Can I enlarge a small image with Playground AI MCP?** Yes, use the `upscale_image` tool. You can pass an existing image URL and specify whether you want a 2x or 4x increase in resolution. **What is the best way to change an image's style using Playground AI MCP?** Use `transform_image`. You provide your public image URL, a text prompt describing the new look, and a strength value (0-1) to control how much the original photo changes. **Does Playground AI MCP support background removal?** Yes, the `remove_background` tool automatically segments the subject from its surroundings, giving you a high-quality, transparent PNG asset. **How do I expand an image's scene using Playground AI MCP?** Use `outpaint_image`. You pass your original picture URL and specify the direction (up/down/left/right) to instruct the model on generating new context.