KreadoAI MCP & Skills Documentation
1. Quick Start
Welcome to the KreadoAI MCP & Skills Documentation! 🙌
This guide will help you quickly integrate KreadoAI's MCP services and Skills capabilities. It covers authentication, MCP architecture, API tools, usage examples, and best practices for building AI-powered workflows.
Whether you're generating marketing assets, creating digital human videos, or building AI agents, this documentation provides everything you need to get started.
👉 Submit your integration requirements here, and our team will contact you within 24 business hours: https://tally.so/r/mJOoeK
1.1 Create an Account
Create a KreadoAI account and upgrade to a Professional Plan.
With KreadoAI's powerful AI capabilities, you can create:
Digital human videos
AI-generated marketing assets
AI voiceovers
AI image and video content
Automated content production workflows
1.2 Obtain Your API Key
After logging into your KreadoAI account, navigate to:
Account Overview → API Settings
to generate and manage your API credentials.
1.3 Important Notes
Temporary request failures may occur during service upgrades or maintenance periods. We recommend implementing a retry mechanism.
Please control request concurrency appropriately. If rate limits are triggered, avoid continuous retries, as this may prolong the cooldown period.
2. MCP Endpoint
To access KreadoAI MCP services, all requests must be authenticated using HTTP headers.
MCP Configuration Example
Supported MCP Clients
KreadoAI MCP can be integrated with:
Cursor
Claude
Custom AI Agens
Any MCP-compatible client
3. AI Image Generation
Use Cases
Generate:
Advertising creatives
Marketing posters
Product images
Brand visuals
E-commerce product listings
Social media assets
Supports:
Text-to-Image (T2I)
Image-to-Image (I2I)
Reference images can be provided using:
Image URL
Base64-encoded image
Supported models:
GPT Image
Gemini
Doubao Seedream
MCP Tools
3.1 get_my_user_detail
Retrieves detailed information for the currently authenticated user.
Returns:
Account information
Email address
Phone number
Membership level
Remaining K Coins
Available quotas
Registration source
Notes:
Authentication is based on the current session context.
userIdcannot be provided manually.Accessing other users' information is not supported.
3.2 submit_ai_image_task
Creates an AI image generation task and returns a task ID.
Supported features:
Text-to-Image generation
Image-to-Image generation
URL reference images
Base64 reference images
3.3 get_ai_image_task_detail
Retrieves task execution results.
Returns:
1
Pending
2
Processing
3
Completed
4
Failed
Additional fields:
Generated image URL
Prompt
Model provider
3.4 resubmit_ai_image_task
Recreates an image generation task using the original task parameters.
Returns a new task ID.
Only tasks owned by the current user may be regenerated.
4. Multi-Size Ad Creative Generation
Use Cases
KreadoAI AI Ad Creative Batch Image Generation allows you to upload product images and specify a list of target sizes to generate multiple ad creatives in a single batch.
It supports common advertising asset sizes for platforms such as Douyin, Kuaishou, Xiaohongshu, Facebook, Google, and TikTok.
Three layout modes are supported:
Original Image Layout
AI Smart Layout
Custom Layout
Multiple output sizes can be generated in a single request.
Important Notes
Request rate: Not specified.
Usage limits: See the tool-specific limitations below.
MCP Tools
Tool
Description
submit_ai_ad_creative_batch_image
Submit a batch image generation task with one or more target sizes.
get_ai_ad_creative_batch_image_detail
Query the status and generated results of a batch image generation task.
submit_ai_ad_creative_batch_image
Submits an AI ad creative batch image generation task. Upload product images and specify one or more target sizes to generate multiple ad creative formats.
Parameters
Parameter
Required
Type
Description
fileUrlList
Yes
FileUrlReq[]
Product images. fileName is used as the basis for naming generated output files. Meaningful file names are recommended.
imageSizeList
Yes
AiBatchImageSizeReq[]
Target output sizes. Supports 1–9 items. Each item must exactly match a size combination listed in the size tables below. Custom size combinations are not allowed.
layout
No
int
1 = Original Image Layout / 2 = AI Smart Layout / 3 = Custom Layout
referenceFileList
No
FileUrlReq[]
Reference images for the custom layout. Required only when layout=3.
dpi
No
string
1k / 2k / 4k. Pixel dimensions must exactly match the selected resolution tier in the size tables below.
File Object: FileUrlReq
Field
Description
fileSource
1 = KreadoAI File Library / 2 = Third-Party URL / 3 = Base64
fileId
Required when fileSource=1.
fileUrl
Required when fileSource=2 or 3.
fileName
Recommended. Used as the basis for naming generated output files.
thumbnailFileUrl
Optional.
Size Object: AiBatchImageSizeReq
All three fields are required for each item. The width and height values must exactly match the size table for the selected dpi.
Example Requests
Product Image from the File Library
Generate an ultra-wide landscape image and a wide landscape image at 1k resolution using the Original Image Layout:
Product Image from a Third-Party URL with a Custom Layout Reference
Important Notes
Maximum 9 output sizes per request. Requests containing more than 9 sizes are rejected by the frontend and server.
Do not calculate pixel dimensions by scaling the aspect ratio. The dimensions for the same
sizeRatioare not always simple multiples across1k,2k, and4k. For example,9:16is1080 × 1920at1kand1440 × 2560at2k. Always use the exact dimensions listed in the corresponding size table.dpiandimageSizeListmust use the same resolution tier. For example, whendpiis2k, you cannot submit the1kdimensions1200 × 1200.When
layout=3,referenceFileListis required. It is not required forlayout=1orlayout=2.Do not include
createUserId,requestSource, ortaskIdwhen submitting a task.
Size Tables
Important: The width and height values must be copied exactly from the corresponding resolution table. Do not calculate or modify them manually.
1K
sizeRatio
Name
width
height
1:1
Square
1200
1200
4:5
Portrait
1200
1500
5:4
Landscape
1500
1200
9:16
Portrait
1080
1920
16:9
Landscape
1920
1080
1.91:1
Landscape
1200
628
4:3
Landscape
1024
768
3:4
Portrait
768
1024
3:2
Landscape
1200
800
2:3
Portrait
800
1200
2:1
Ultra-wide Landscape
1200
600
1:2
Portrait
632
1264
6:5
Near-square
1200
1000
1:8
Extra-tall Portrait
352
2928
8:1
Ultra-wide Landscape
2928
352
1:4
Tall Portrait
512
2064
4:1
Wide Landscape
2064
512
4:15
Tall Portrait
512
1920
32:5
Ultra-wide Landscape
2432
380
97:25
Wide Landscape
1940
500
21:9
Ultra-wide Landscape
1568
672
364:45
Ultra-wide Landscape
2912
360
2K
sizeRatio
width
height
1:1
2048
2048
4:5
1800
2250
5:4
2250
1800
9:16
1440
2560
16:9
2560
1440
1.91:1
2048
1072
4:3
1920
1440
3:4
1440
1920
3:2
2400
1600
2:3
1600
2400
2:1
2400
1200
1:2
1264
2528
6:5
1800
1500
1:8
704
5856
8:1
5856
704
1:4
1024
4128
4:1
4128
1024
4:15
1024
3840
32:5
4864
760
97:25
3880
1000
21:9
3136
1344
364:45
5824
720
4k
sizeRatio
width
height
1:1
4096
4096
4:5
3600
4500
5:4
4500
3600
9:16
2160
3840
16:9
3840
2160
1.91:1
4096
2144
4:3
3840
2880
3:4
2880
3840
3:2
4000
2667
2:3
2667
4000
2:1
4000
2000
1:2
2528
5056
6:5
3840
3200
1:8
1408
11712
8:1
11712
1408
1:4
2048
8256
4:1
8256
2048
4:15
2048
7680
32:5
9728
1520
97:25
7760
2000
21:9
6272
2688
364:45
11648
1440
get_ai_ad_creative_batch_image_detail
Retrieves the execution details and generated results of an AI ad creative batch image generation task.
Use the task ID returned by submit_ai_ad_creative_batch_image to query the task status and generated assets.
Response
The response includes:
Field / Information
Description
status
Task status: 1 = Pending, 2 = Processing, 3 = Success, 4 = Failed.
Subtask list
A list of subtasks corresponding to each product image.
Generated file URL
URL of each generated image.
Generated thumbnail URL
Thumbnail URL for each generated image.
Dimensions
Width and height of each generated image.
Model information
The model/source used to generate each asset.
When the task succeeds, the response contains the generated image files and their corresponding metadata.
5. Website Data Crawling (For AI Ad Generation)
Use Cases
Crawl and analyze:
Brand websites
Landing pages
E-commerce product pages
AI extracts structured product and brand information for downstream AI content generation.
Processing time:
Typically 1–5 minutes
Important Notes
Recommended polling interval: 10–15 seconds
Maximum waiting time: 15 minutes
Result TTL: 24 hours
During polling, set:
After task completion:
Business source:
1
AI Marketing Video (Default)
2
AI Advertising Image
MCP Tools
5.1 submit_crawl_task
Submits a webpage crawling task.
Parameters
query
Yes
Target webpage URL
source
No
Business source
sessionId
No
Session identifier
Returns:
5.2 get_crawl_task_status
Retrieves task status and crawling results.
Parameters
threadId
Yes
includeContent
No
Status
RUNNING
Processing
SUCCESS
Completed
FAILED
Failed
NOT_FOUND
Not Found / Expired
Response Fields
threadId
sessionId
status
currentNode
crawlId
url
source
errorMsg
content
Content Fields
title
productName
brandName
price
currency
productDescription
productFeatures
targetAudience
useScenario
imageUrls
videoUrls
logoUrl
language
location
productLllmResult
metaDescription
content
6. AI Advertising Image Generation
Use Cases
Generate marketing images directly from:
Product URLs
Product content
Supported modes:
generationMode=1
AI Creative Generation
generationMode=2
Template-Based Generation
Customizable:
Prompt
Aspect ratio
Resolution
Logo
Brand color
MCP Tools
6.1 submit_url2image_task
Creates AI advertising image generation tasks.
Returns:
6.2 get_ai_ad_creative_batch_image_detail
Retrieves task execution results and generated image URLs.
7. Text-to-Speech (TTS)
Use Cases
Convert text into MP3 audio for:
Digital human videos
Advertising assets
Marketing campaigns
Recommended workflow:
Important Notes
Maximum 1 request per second
Valid apiToken required
voiceSource is required when voiceClone=0
MCP Tools
7.1 get_voice_language_list
Returns supported languages.
Response:
7.2 get_voice_list
Returns voice models matching specified filters.
Supported providers:
1
Microsoft
3
Alibaba
4
ByteDance
5
MiniMax
6
21
ElevenLabs
Response:
7.3 text_to_speech
Converts text into MP3 audio.
Returns:
taskId
audioUrl
fileId
duration
paymentMoney
Response:
8. Digital Human Video
Use Cases
Generate lip-synced marketing videos using:
Real-person avatars
Photo avatars
Cloned digital humans
Recommended workflow:
MCP Tools
8.1 get_digital_human_avatar_list
Returns available digital human avatars.
Response:
8.2 submit_system_lip_task
Creates a lip-sync generation task.
Returns:
Contains:
8.3 get_lip_video_result
Retrieves lip-sync generation results.
Status:
1
Pending
2
Processing
3
Completed
4
Failed
5
Timeout
Response:
9. AI Video Generation (Image-to-Video / Text-to-Video)
Use Cases
Generate marketing videos using:
Text prompts
Images
Multi-image references
Multimodal references
Video continuation
Supported model families:
Doubao Seedance Series
Veo
Sora
Vidu
Kling
Happy Horse
Wan
MiniMax
Workflow
MCP Tools
9.1 submit_image_to_video_task
Creates image-to-video or text-to-video generation tasks.
Returns:
Includes:
taskId
taskIds
Supported Reference Modes
none
first-last
reference
multimodal-reference
video-continuation
Supported Media Types
image
video
audio
9.2 batch_get_image_to_video_detail
Retrieves generation results.
Status:
1
Pending / Processing
2
Success
3
Failed
4
Failed
Response:
10. Subtitle & Watermark Removal
Use Cases
Remove:
Video subtitles
Watermarks
Required inputs:
Source video URL
Subtitle coordinates
Watermark coordinates
MCP Tools
10.1 submit_subtitle_removal_task
Creates a subtitle or watermark removal task.
Returns:
10.2 get_subtitle_removal_result
Retrieves task results.
Status:
1
Pending
2
Processing
3
Completed
4
Failed
5
Timeout
Returns:
11. Smart Canvas
This chapter covers the KreadoAI Smart Canvas MCP tools. It starts with the basic tools for managing projects and querying tasks, then introduces task-specific submission tools. Background removal is the only supported task in this release.
Use Cases
KreadoAI Smart Canvas allows you to submit image processing tasks to the current user's canvas project and returns a
batchIdandtaskIds.Projects can be created or listed using the basic tools. For task submission,
projectIdcan also be omitted. In that case, the service automatically creates a project named using the current time in theyyyyMMdd-HHmmssformat (Asia/Shanghai).The only supported task in this release is background removal.
Task submission is asynchronous and does not wait for completion. You must call
mcp_query_canvas_task_resultsto poll the task status until it reaches a terminal state.
Important Notes
Authentication: Authentication relies on the currently logged-in user. If the user is not logged in, the request is rejected and the canvas service is not called.
Public Tool names: All four tools use the
mcp_*naming convention.Projects:
mcp_create_canvas_projectandmcp_list_canvas_projectsare shared across all task types. For background removal,projectIdcan be omitted to automatically create a project. If a non-empty but invalid or unauthorizedprojectIdis provided, the request returns an error instead of creating a new project.Reuse existing projects (recommended): If a reusable project is available—for example, a
projectIdreturned by a previous submission or a project listed bymcp_list_canvas_projects—pass thatprojectIddirectly when submitting a task. Do not callmcp_create_canvas_projectagain, as each call creates a new project.Polling (required): After any successful task submission, call
mcp_query_canvas_task_results. Continue polling whenstatusis0,1, or2. Whenstatusis3, read the result URL. Whenstatusis4or5, readerrorMsg. Stop polling once a terminal status is reached.
Overview
Choose the Right Tool
Your goal
Tool
Notes
Create a canvas project
mcp_create_canvas_project
Shared across all task types
List existing projects and retrieve a valid projectId
mcp_list_canvas_projects
Shared across all task types
Query task progress and results
mcp_query_canvas_task_results
Shared across all submitted tasks
Batch background removal
mcp_image_remove_bg
Available in this release. Poll for results after submission. Results are typically available within 3–5 seconds.
Tool Reference (4 Tools)
#
Tool
Category
Description
1
mcp_create_canvas_project
Basic
Create a canvas project for the current user
2
mcp_list_canvas_projects
Basic
List non-deleted canvas projects for the current user, with pagination
3
mcp_query_canvas_task_results
Basic
Query task details by batchId and/or taskIds
4
mcp_image_remove_bg
Task: Background Removal
Remove backgrounds from multiple images
Basic Tools
1. mcp_create_canvas_project
Creates a Smart Canvas project for the current user.
Parameters
Parameter
Required
Description
name
No
Project name. After trimming, the name must not exceed 100 characters. If the name exceeds this limit, the server rejects the request without truncating it. If omitted or blank, the service uses the current time in the yyyyMMdd-HHmmss format (Asia/Shanghai), for example, 20260909-135152.
Response
Field
Description
id
ID of the newly created project. Use this value as projectId in subsequent task submission calls.
name
The name saved for the project: either the specified name or the automatically generated time-based name.
Example Request
2. mcp_list_canvas_projects
Lists all non-deleted Smart Canvas projects belonging to the current user. Use this tool to retrieve a valid projectId.
Parameters
Parameter
Required
Description
name
No
Fuzzy filter by project name. If omitted, all projects are listed.
current
No
Page number. Defaults to 1. Values less than 1 are treated as 1.
size
No
Number of items per page. Defaults to 50. Values less than 1 are treated as 50; values greater than 100 are treated as 100.
Response
Field
Description
total
Total number of matching projects.
current
Actual page number returned.
size
Actual page size returned.
items
List of projects.
items[] Fields
Field
Description
id
Project ID.
name
Project name.
createTime
Creation time, returned when available. Format: yyyy-MM-dd HH:mm:ss.
updateTime
Last update time, returned when available. Format: yyyy-MM-dd HH:mm:ss.
Example Request
3. mcp_query_canvas_task_results
Queries task details for the current user by batchId and/or taskIds.
At least one of batchId or taskIds must be provided. If both are empty, the canvas service is not called.
Parameters
Parameter
Required
Description
batchId
Conditional
Batch ID returned by a task submission. At least one of batchId or taskIds must be provided.
taskIds
Conditional
List of task IDs. At least one of batchId or taskIds must be provided.
Response
Returns an array of task details. Each item contains the following fields:
Field
Description
taskId
Task ID.
batchId
Batch ID.
layerId
System-generated layer ID. Generally not needed by callers.
parentLayerId
Parent layer ID. Generally not needed by callers.
taskType
Numeric task type. Background removal is 5.
taskTypeValue
String task type. Background removal is image_remove_bg.
resultType
Result type: 1 = image, 2 = video, 3 = text. Background removal returns 1.
resultTypeValue
Result type as a string. Background removal returns image.
status
Task status: 0 = Pending submission, 1 = Submitted, 2 = Processing, 3 = Success, 4 = Failed, 5 = Timed out. Statuses 3, 4, and 5 are terminal states. Stop polling when one of these states is reached.
errorMsg
Error details. Usually empty when the task succeeds.
resultFileId
ID of the successfully generated result file. Can be used as fileId in subsequent submissions together with resultFileUrl.
resultFileUrl
URL of the successfully generated result file.
resultThumbnailFileUrl
URL of the result thumbnail.
innerResultData
Extended result data map. Contents depend on the task type. Callers generally do not need to parse this field.
Example Request
Background Removal
Removes the background from one or more images.
Supported inputs include a KreadoAI fileId, a KreadoAI file URL, an external URL, or image base64 data. If no fileId is provided, the image is uploaded to internal storage.
imagesmust contain at least 1 and at most 50 items. If any input asset fails validation or processing, the entire batch submission is rejected.When
fileIdis provided, the service resolves the file URL from the file library and does not re-upload the file. When only an HTTP(S) URL or base64 data is provided, the service uploads the image to internal storage.A project is created automatically only when
projectIdis omitted or blank.
4. mcp_image_remove_bg
Submits a batch background removal task to Smart Canvas. A single image is also submitted as a batch containing one task.
Parameters
Parameter
Required
Description
projectId
No
ID of an existing project. Pass a reusable project ID whenever possible to avoid creating a new project for every submission. If omitted or blank, the service automatically creates a project using the same naming rules as mcp_create_canvas_project, with a default name in the yyyyMMdd-HHmmss format. The response returns projectCreated=true. If a non-empty but invalid or unauthorized projectId is provided, the request returns an error instead of creating a new project.
images
Yes
List of images to process. Must contain at least 1 and at most 50 items. See the table below for the item structure.
images[] Fields
Field
Required
Description
fileId
Conditional
ID of a file in the KreadoAI file library (b_file.id). When provided, the service resolves the URL from the file library and does not re-upload the file.
imageUrl
Conditional
Image URL (http/https) or image base64 data. Required when fileId is not provided. Each item must contain at least one of fileId or imageUrl. External URLs and KreadoAI CDN URLs are uploaded to internal storage.
Response
Field
Description
batchId
Batch ID used to poll task results.
taskIds
List of task IDs in the batch.
projectId
ID of the project used for the submission.
projectCreated
Whether a new project was created for this submission: true or false.
taskTypeValue
Always image_remove_bg.
batchTotal
Number of tasks in the batch.
pollHint
Indicates that mcp_query_canvas_task_results should be called to poll for results. This submission call does not wait for completion.
Example Request (Mixing fileId and URL)
Example Request (URL Only)
Common Errors
Issue
How to Handle It
Both fileId and imageUrl are empty
Each item in images must contain at least one of these fields.
File not found or unauthorized fileId
Use a file belonging to the current user, or provide an accessible URL instead.
images is empty or contains more than 50 items
Provide between 1 and 50 images per request.
Invalid projectId
Use a valid project ID belonging to the current user, or omit it to create a project automatically.
name exceeds 100 characters
Shorten the project name. The server does not truncate it.
Both batchId and taskIds are empty
Provide at least one of these fields.
Canvas service unavailable
Retry later. Do not submit the same task again to avoid duplicate charges.
12. Video Quality Enhancement
Use Cases
KreadoAI Video Quality Enhancement improves the clarity and visual quality of input videos and returns a task ID. After successful submission, call
get_video_enhance_detailto poll for the result.Input videos can be provided using a system file library
fileIdor a publicly accessible third-party URL. Third-party URLs are first uploaded to internal storage. Base64-encoded videos are not supported.Enhancement types:
enhanceType:1= Standard,2= AI Model
resolution:720p,1080p, or2k(supported by both enhancement types)
fps: Integer between15and120, inclusive. Common values include24,30,45,60, and120.
scene: Only applicable to the Standard enhancement type. Defaults tocommonif omitted. Ignored for the AI Model enhancement type.Input videos must be between 3 seconds and 5 minutes long.
Important Notes
Request rate limit: You can submit a maximum of 1 task per second.
MCP Tool Reference
submit_video_enhance_task (Submit Task)
Enhances the clarity and visual quality of an input video and returns a task ID. After submission, call get_video_enhance_detail to poll for the result.
Parameters
Parameter
Required
Description
video
Yes
Input video object. See the table below for its fields.
enhanceType
Yes
Enhancement type: 1 = Standard, 2 = AI Model.
resolution
Yes
Target resolution: 720p, 1080p, or 2k. Supported by both enhancement types. 4k is rejected.
fps
Yes
Target frame rate. Must be an integer between 15 and 120, inclusive. Common values include 24, 30, 45, 60, and 120.
scene
No
Scene type. Only applicable to the Standard enhancement type. Defaults to common if omitted. Ignored for the AI Model enhancement type.
batchId
No
Batch UUID. Use the same value when submitting multiple videos in one batch. If omitted, the server generates one.
video Object Fields
Field
Required
Description
fileId
One of two
ID of a file in the system file library (b_file.id). If both fileId and url are provided, fileId takes precedence.
url
One of two
Publicly accessible third-party video URL (http or https). The video is first uploaded to internal storage before submission. Base64-encoded videos are not supported.
Validation Rules
The input video must be between 3 seconds and 5 minutes long. Requests that fall outside this range are rejected immediately and do not incur any charges.
Response
Returns:
Field
Description
taskId
Task ID as a string.
get_video_enhance_detail (Query Task Details)
Retrieves the details and processing status of a video quality enhancement task.
Parameters
Parameter
Required
Description
taskId
Yes
Task ID returned by submit_video_enhance_task.
Response Fields
Field
Description
id / batchId
Task ID / Batch identifier.
status
Task status: 1 = Pending, 2 = Processing, 3 = Success, 4 = Failed.
enhanceType / scene / resolution / fps
Echoes the submitted parameters.
sourceFile
Input asset JSON, including fileId, internal URL, and thumbnail.
duration
Input video duration in whole seconds. Used as the billing basis.
deductAmount
Actual amount deducted, in K-Coins.
resultFile
Result asset JSON, including fileId, internal URL, dimensions, duration, and thumbnail. Populated when the task succeeds.
error
Reason for failure.
createTime / updateTime
Creation and last update timestamps.
Access Restrictions
You can only query your own tasks. Requests using another user's taskId are rejected.
Contact Us
Need technical support or want to learn more about KreadoAI MCP?
📧 Email: support@kreadoai.com
🌐 Website: www.kreadoai.com
Our team is always available to assist with your integration and deployment needs.
Last updated