# Midjourney MCP > Midjourney AI (Generative Image Arts) lets you generate professional, high-fidelity artwork directly from your agent. Use text prompts for new scenes, or feed it existing images to reverse engineer concepts and create variations. It handles complex visual tasks like blending multiple sources, upscaling specific parts of a grid, and simulating camera movements, giving designers full control over their artistic output. ## Overview - **Category:** superpower - **Price:** Free - **Tags:** generative-art, text-to-image, ai-design, creative-tools, upscaling ## Description This MCP gives you complete command over Midjourney's generative art tools, letting your agent handle the heavy lifting of professional image creation. Instead of wrestling with complex web interfaces or endless manual prompting, you talk to it conversationally and get results. You can generate entire scenes from a simple text description, but you don't stop there. If you love a particular angle in an initial 2x2 grid, you can pull out just that tile for high-resolution upscaling. Need to iterate on a style? Just tell your agent to make variations or blend two completely different source images into one cohesive piece. The whole process, from generating the concept to refining the final asset, runs through your preferred AI client and is managed by Vinkius, making advanced creative tools feel simple. You just describe what you need, and it manages the job status until the image is ready. ## Tools ### generate_image Creates a new image based on a text prompt using Midjourney's generative models. ### get_job Checks the current status of an ongoing creative job to see if it’s pending or complete. ### list_jobs Retrieves a history of all your recently executed Midjourney prompts and jobs. ### describe_image Analyzes an image URL and returns four potential text descriptions that might have created it. ### blend_images Combines two to five specific images into a single, unique artistic composition. ### upscale_image Selects and increases the resolution of one specific tile from an image grid. ### generate_variation Creates new, structural artistic options based on a single existing image in a grid. ### reroll_job Re-runs an identical prompt exactly as before to generate a completely new set of outputs. ### pan_image Extrapolates the image borders by simulating a camera movement in a specific direction (up, down, left, or right). ### zoom_out_image Widens the visible perspective of an existing scene by zooming out. ## Prompt Examples **Prompt:** ``` Generate a futuristic cyberpunk city at sunset with neon reflections in 16:9 ``` **Response:** ``` Generating image… I've triggered the Midjourney imagine task with your prompt and a 16:9 aspect ratio. Your Job ID is 'mj-98765'. I'm monitoring the process. High-fidelity textures and lighting are being computed. I'll provide the 2x2 grid URL as soon as it's ready. ``` **Prompt:** ``` Upscale the first image from grid ID 'mj-98765' ``` **Response:** ``` Upscaling tile 1… I've triggered the high-resolution render for your first choice. This will isolate the image and apply advanced AI detail enhancement. I'll provide the final high-res URL once the upscale job completes. ``` **Prompt:** ``` Describe the contents of this image: [url] ``` **Response:** ``` Analyzing visual concepts… I've retrieved 4 candidate prompts from Midjourney: 1) 'Minimalist architectural sketch in charcoal', 2) 'Brutalist concrete building perspective', 3) 'Abstract geometric shadows on stone', and 4) 'Architectural blueprint style'. Would you like to use any of these to generate a new variation? ``` ## Capabilities ### Generate images from text You provide a description, and the agent triggers Midjourney to create a brand new piece of art. ### Refine existing visual concepts The MCP creates structural variations or entirely new versions based on an image you already generated. ### Blend multiple source images You give it a handful of pictures, and the agent merges them into one cohesive, unique composition. ### Enhance resolution or zoom The MCP takes specific tiles from an image grid and professionally upscales them, or simulates camera movements like panning or zooming out a scene. ### Determine source prompts You provide a public image URL, and the agent analyzes it to return four suggested text descriptions of what's in the picture. ## Use Cases ### Creating a cinematic moodboard A director needs a set of background images for a sci-fi film. They prompt their agent with 'futuristic city at sunset' (`generate_image`). After getting the 2x2 grid, they use `pan_image` to simulate a sweeping view and then call `blend_images` using three separate source photos to ground the art in reality. ### Improving product mockups A marketing team generates an initial image of a shoe (`generate_image`). They realize the detail needs work, so they use `upscale_image` on the sole tile. Then, if that doesn't look right, they run `reroll_job` to generate another set of options. ### Developing visual storyboards A game designer needs a sequence of shots. They start with an image and use the agent to call `zoom_out_image`, then follow up by calling `generate_variation` on that new, wider frame, ensuring consistency across multiple 'shots'. ### Reverse-engineering artistic styles A designer finds a beautiful piece of concept art online. They feed the URL to the agent and call `describe_image`, which gives them four candidate prompts they can then use in `generate_variation` to replicate the style. ## Benefits - You bypass manual prompt entry. Simply tell your agent what you want, and it handles the entire generation process using `generate_image`. - Need to improve a single detail? Instead of accepting the initial render, use `upscale_image` to isolate and enhance only the tile you like best. - Concepting is easier than ever. If you see an image online but don't know how it was made, run `describe_image` on its URL to pull out potential source prompts. - Don't settle for one look. Use `generate_variation` or the `blend_images` tool to quickly iterate and combine styles until your vision is perfect. - Manage complexity with job tracking. The MCP handles status updates automatically; you just ask it to check progress using `get_job`. ## How It Works The bottom line is that your AI client manages every step of the complex creative process so you never have to manually check job statuses or piece together disparate images. 1. Subscribe to this MCP on Vinkius and input your Midjourney API key. 2. Tell your AI client which creative task you need, such as generating a scene or blending images. The agent executes the command and retrieves an initial Job ID. 3. Your agent monitors the job status automatically. When finished, it hands you the final high-resolution image or data package. ## Frequently Asked Questions **How do I check if my Midjourney job is done using the Midjourney AI (Generative Image Arts) MCP?** You use `get_job` to check the status of any running task. Your agent monitors the process, so you just need to ask it for an update and provide the current Job ID. **Can Midjourney AI (Generative Image Arts) create art from a drawing I find online?** Yes. You send the image URL to `describe_image`, and the tool will return four candidate text descriptions, helping you understand what prompt was used. **What is the difference between generate_variation and reroll_job in Midjourney AI (Generative Image Arts)?** `generate_variation` creates new structural options based on a specific image within a grid. `reroll_job`, however, re-runs the exact same prompt arguments to give you an entirely fresh set of results. **Does this MCP support blending photos from different sources?** Yes, you can use `blend_images` to merge two to five explicit source images into a single unique composition, bridging different artistic styles flawlessly.