BagelImageGeneration
Open-source multimodal model from ByteDance for image generation, combining understanding and generation capabilities with advanced editing features
LABSCO SUMMARY
The named formats are prompt engineering someone already did for you — and they are also why this is expensive to leave connected.
"Make a knowledge card" reliably produces a knowledge card, where the same request as a raw prompt produces a picture of a card. That packaging is genuinely useful. The trade is size: this is one of the larger tool sets in the catalogue, and all of it rides along on every turn. If you only want plain generation, their custom server lets you take a subset instead.
What it is
302AI's image server, notable for how many named output formats it exposes rather than leaving everything to the prompt.
What you get
- General generation and editing across several underlying models
- Named formats — posters, stickers, knowledge cards, quotation cards, comic and brick-built styles, graffiti
- Fetch tools for retrieving results from the longer-running model backends
Requirements
A 302AI API key, and npx. MIT licensed; the vendor's own package.
Setup effort
One command plus a key — npx -y @302ai/image-mcp, then supply credentials
USE CASES
01
Generating a picture
Straight generation across several model backends.
02
Producing a specific artefact
Named formats rather than prompt-wrangling.
03
Editing what you have
Edit rather than regenerate.
54 tools
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
Deblur
Removes motion blur from photos using AI, enhancing image clarity and restoring sharp details.
Denoise
Removes color noise from photos using AI, enhancing image quality and clarity.
Erase_Foreground
Input an image URL to intelligently identify and remove the main foreground object. The tool then naturally fills in the background using AI technology. Ideal for scenarios such as removing unwanted people or objects, or creating clean background images.
Eraser
Input the original image URL and the mask image URL. The tool will erase the specified object in the image based on the mask (white area) and intelligently fill the removed area using AI technology to seamlessly blend with the surrounding environment.
Fetch_Hunyuan3d
Query the result of asynchronous Hunyuan3d task by task ID. If the task is not completed, return the request ID so that querying can continue.
Fetch_Images_expand
Query the result of asynchronous Images-expand task by task ID. If the task is not completed, return the request ID so that querying can continue.
Fetch_Kling
Query the result of an asynchronous Kling task by task ID. If the status is not 'SUCCESS', the request ID will be returned for continued querying.
Fetch_Portrait
Portrait photo result retrieval: Use the task_id and req_json from the previous task to query. If the result is still being generated, task_id and req_json will be returned and you should query again.
Fetch_Reference2Image
Query the result of an asynchronous Reference2Image task by task ID. If the status is not 'SUCCESS', the request ID will be returned for continued querying.
Generate_Background
Input image URL and corresponding parameters to intelligently replace or generate new backgrounds, seamlessly blending original subjects with AI-generated backgrounds. Suitable for product photography, portrait processing, creative design and other scenarios.
GetTask
Retrieve the results of Qwen-MT-Image and Wanx2.1-ImageEdit tasks. If no results are available, clearly return the task_id to the user for use in subsequent requests
gpt4oImageEdit
Edit an existing image using a text prompt with the gpt-image-1 model.
gpt4oImageGeneration
Generate images based on text prompts using the gpt-image-1 model.
HidreamImageGenerationDev
Edit images based on input URL and prompt
HidreamImageGenerationFast
Edit images based on input URL and prompt
HidreamImageGenerationFull
Edit images based on input URL and prompt
HtmltoPng
Converts HTML code into a PNG image based on the input string.
Hunyuan3d
Convert the input image URL into a 3D model. This interface is asynchronous: after submitting the task, obtain request_id and retrieve results using Fetch_Hunyuan3d.
imageGenerate
Generate images from text prompts using various AI models. Supports multiple model providers including DALL-E, Stability AI, Flux, Ideogram, Google, and more. You can specify parameters like model type, dimensions, negative prompts, and aspect ratio. Some models support image-to-image generation by providing an input image.
Images-expand
Image extension feature from Kling, input URL and prompts to perform image extension, supports various parameters. The interface uses asynchronous submission and requires querying results through Fetch_Images_expand MCP
iRAGImageGeneration
Baidu iRAG Image Generation Model
Kling_Images_Generations
This tool supports both Text-to-Image and Image-to-Image generation. In Text-to-Image mode, provide only a text prompt to generate images. In Image-to-Image mode, you must additionally provide a reference image URL (Base64 or accessible link).
knowledgeCardGenerate
Generate knowledge cards with various styles and content. Supports customization of style, date, content, language, and QR codes. Based on https://302.ai/tools/card/
LinktoIMG
Convert the input link into an image and return it as a URL.
MinimaxImageGeneration
MiniMax Image-01 Generation Model
Moondream2
Analyzes the content of the provided image URL and generates a descriptive prompt for downstream AI tasks.
Nano-Banana-Image Edit
Edit images based on input URL and prompt
Nano-Banana-Image Generation
Nano-Banana text-to-image generation
Nano-Banana-Pro-Image Edit
Edit images based on input URL and prompt
Nano-Banana-Pro-Image Generation
Nano-Banana-Pro text-to-image generation
philosophyCardGenerate
Generate philosophy cards with various styles and content. Supports customization of style, content, and language. Based on https://302.ai/tools/card/
philosophyQuotationGenerate
Generate philosophy quotation cards with various styles. Supports customization of style, quotation content, creator name, and name location. Based on https://302.ai/tools/card/
Portrait
By referencing the facial features of a single real person photo as input, this function uses text prompts to generate a variety of realistic portrait photos. It can be widely used in interactive entertainment, portrait effects, and e-commerce marketing scenarios. Preset style keywords can be used as templates, or you can experiment with your own templates based on examples. This is an asynchronous API; you need to query the interface to get the result.
posterGenerate
Generate posters with various styles and content. Supports customization of style, content, and language. Based on https://302.ai/tools/card/
Qwen-Image
Qwen-Image, a text-to-image model from the Qwen series, is a general-purpose image generation model. It supports various artistic styles and excels at rendering complex text, especially Chinese and English. The model supports multi-line layouts, paragraph-level text generation, and fine-grained detail depiction, enabling complex mixed text-and-image layout designs.
Qwen-MT-Image
Tongyi Qianwen Image Translation Model (Qwen-MT-Image) supports translating text from images in 11 languages into Chinese or English, accurately preserving original layout and content information, with features like terminology definition, sensitive word filtering, and image subject detection
QwenImageEdit
Tongyi Qianwen Image Editing Model (Qwen-Image-Edit) supports precise bilingual text editing in Chinese and English, color adjustment, detail enhancement, style transfer, object addition/removal, position and action changes, enabling complex image-text editing operations.
Reference2Image
Generate images by providing reference image URLs and a prompt. This is an asynchronous API and requires using Fetch_Reference2Image to obtain the result.
Retouch
Reads the image from the provided URL and automatically beautifies facial photos.
SAM
Generates segmentation mask images from input prompts and image URLs. Supports object or semantic segmentation for AI vision tasks.
seedEdit
Use SeedEdit to edit images with precise understanding of both English and Chinese instructions. With simple prompts, you can change backgrounds, transform styles, add/remove objects, and more. Ideal for: 1) Modifying objects or scenes in images; 2) Changing image style or mood; 3) Adding or removing elements from images; 4) Adjusting image atmosphere or environment. Unlike generating completely new images, this tool makes precise modifications while preserving most of the original content.
Seedream4
DoubTech SeedreamⅣ is the latest model supporting both text-to-image and image-to-image generation. Model name: doubao-seedream-4-0-250828. Features include: Sequential Image Generation (set sequential_image_generation to 'auto') - Multi-image to sequential: Generate a series of related images from 2-10 reference images + text prompt (total input + output ≤ 15 images); Single-image to sequential: Generate up to 14 related images from 1 reference image + text prompt; Text to sequential: Generate up to 15 related images from text prompt only. Single Image Generation (set sequential_image_generation to 'disabled') - Multi-image to single: Generate 1 image from 2-10 reference images + text prompt; Single-image to single: Generate 1 image from 1 reference image + text prompt; Text to single: Generate 1 image from text prompt only.
soul
This tool supports both Text-to-Image and Image-to-Image generation. In Text-to-Image mode, provide only a text prompt to generate images. In Image-to-Image mode, you must additionally provide a reference image URL (Base64 or accessible link).
SvgToPng
Converts SVG code into PNG image format based on the input string.
TextToGraffiti
Pass in a description to generate a doodle image
TextToImage
Pass in a description to generate an image
TextToLego
Pass in a description to generate a LEGO-style image
TextToSticker
Pass in a description to generate a sticker image
TextToWojakComic
Pass in a description to generate a Wojak comic
Trellis
Converts a 2D image into a detailed 3D model based on the input URL, supporting customizable settings for mesh and texture quality.
Virtual-Tryon
Provides AI-powered virtual clothing try-on. Users can preview how garments look on human figures by uploading their photos and clothing images, producing realistic fitting effects suited for e-commerce, fashion design, and social sharing.
Wanx2.1-ImageEdit
Tongyi Wanxiang Image Editing Model (Wanx2.1-ImageEdit) enables diverse image editing through simple instructions, suitable for image expansion, watermark removal, style transfer, image restoration, and image enhancement scenarios
ZhiPuImageGeneration
Zhipu CogView-4 Image Generation Model
BEFORE YOU INSTALL
What to know first
Tools you get
54 tools when we connected. Six tools, GetTask and the Fetch_* tools, only retrieve results of jobs started by other tools, leaving 48 that do the work. Some names contain spaces, such as Nano-Banana-Pro-Image Generation.
Requires
- npx on PATH
- 302AI_API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
CONTEXT COST
Heavy Standing cost
Rated by the size of the 54 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Connect your client
Server setup
TERMINAL
npx -y @302ai/image-mcpMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"302ai-image-mcp": {
"command": "npx",
"args": [
"-y",
"@302ai/image-mcp"
],
"env": {
"302AI_API_KEY": "<YOUR_AI_API_KEY>"
}
}
}
}
