IMPORTANT ASYNCHRONOUS TASK WORKFLOW
- Call this MCP generation tool once and read taskId from structuredContent.
- Do not call the generation tool again while the task is PENDING or RUNNING.
- Wait at least 5 seconds, then call nexscope_media_task_result with skillSlug set to image-captioning-api, the exact taskId, and the same bearer API key. While the task remains PENDING or RUNNING, use 5, 10, 20, and 30-second progressive backoff. Never query more often than once every 5 seconds.
- Stop querying when the task succeeds or fails. On success, return the final Nexscope resource URLs and the returned count as the actual charged credits.
- Never invent a taskId and never expose internal provider names or provider resource URLs.
Multimodal AI
Image Captioning API MCP Tool
Extract descriptive prompts or selling-point text from one or more images.
MCP tool name
nexscope_image_captioning_apiEquivalent REST API path
https://claw-callback.nexscope.ai/api/skill-api/v1/skills/image-captioning-api/runJSON-RPC method
tools/callAuthentication
Send the user API key as a bearer token on every MCP JSON-RPC request.
Authorization: Bearer nk_xxxxxxxxxxxxxxxxxIf authentication is missing or invalid, ask the user to open API Access and copy a valid key before calling this tool.
Call this MCP tool
Use tools/call with the MCP tool name below. The arguments object should match the table on this page.
{
"jsonrpc": "2.0",
"id": "image-captioning-api-1",
"method": "tools/call",
"params": {
"name": "nexscope_image_captioning_api",
"arguments": {
"type": "product",
"imageUrlList": [
"https://example.com/product.jpg"
],
"keyword": "phone case",
"provider": "default",
"refVideoUrl": "https://example.com/reference-video.mp4",
"saleCountry": "US",
"targetLanguage": "en",
"duration": 15
}
}
}REST API equivalent
POST https://claw-callback.nexscope.ai/api/skill-api/v1/skills/image-captioning-api/runCanonical detail URL
https://www.nexscope.ai/mcp-map/image-captioning-apiAsynchronous task workflow
This capability uses a create-and-query flow. Creating the task and retrieving its final generated resources are two connected requests.
- 1. Create onceCall the MCP tool once and keep the returned
taskId. - 2. Query with backoffCall
nexscope_media_task_resultwith this skill slug and task ID. Wait at least 5 seconds first; while pending or running, back off at 5, 10, 20, then 30-second intervals. - 3. Use final resultStop at a terminal status. Successful results contain Nexscope resource URLs and the actual charged-credit count.
MCP task result call
{
"jsonrpc": "2.0",
"id": "image-captioning-api-result-1",
"method": "tools/call",
"params": {
"name": "nexscope_media_task_result",
"arguments": {
"skillSlug": "image-captioning-api",
"taskId": "TASK_ID_FROM_THE_GENERATION_RESULT"
}
}
}REST fallback
GET https://claw-callback.nexscope.ai/api/skill-api/v1/skills/image-captioning-api/tasks/{taskId}Arguments
| Name | Type | Required | Meaning |
|---|---|---|---|
type | string | required | Extraction type supported by the upstream API. Example: product |
imageUrlList | array | required | Public image URLs used for prompt extraction. Example: [
"https://example.com/product.jpg"
] |
keyword | string | optional | Optional keyword context. Example: phone case |
provider | string | optional | Optional upstream provider code. Example: default |
refVideoUrl | string | optional | Optional reference video URL. Example: https://example.com/reference-video.mp4 |
saleCountry | string | optional | Target sale country code. Example: US |
targetLanguage | string | optional | Target language code for extracted text. Example: en |
duration | integer | optional | Optional target duration in seconds for video-related extraction. Example: 15 |
Response fields
Treat structuredContent as the programmatic API response payload.
| Name | Type | Required | Meaning |
|---|---|---|---|
taskId | string | required | Nexscope asynchronous task id used to query the generation result. Example: 6b3f1c61-1b7d-4a59-8a7d-3b23e4b0c9e2 |
status | string | required | Task status. Initial run returns PENDING; the query endpoint may return PENDING, RUNNING, SUCCESS, FAILED, or TIMEOUT. Example: PENDING |
message | string | optional | User-facing task state message. Example: The task has been created. Use taskId to query the result. |
errcode | integer | optional | Upstream status code returned by the provider. Example: {} |
errmsg | string | optional | Upstream status message returned by the provider. Example: {} |
code | string | optional | Provider-specific status code. Example: {} |
msg | string | optional | Provider-specific status message. Example: {} |
type | string | optional | Provider-specific render or payload type. Example: {} |
title | string | optional | Provider-specific response title. Example: {} |
sourceType | string | optional | Provider-specific source platform type. Example: {} |
sourceTool | string | optional | Provider-specific source tool name. Example: {} |
costToken | integer | optional | Token cost reported by the upstream provider. Example: {} |
costTime | integer | optional | Execution time reported by the upstream provider. Example: {} |
page | integer | optional | Current page returned by the upstream provider. Example: {} |
pageSize | integer | optional | Page size returned by the upstream provider. Example: {} |
pageItemCount | integer | optional | Item count on the current page returned by the upstream provider. Example: {} |
totalPage | integer | optional | Total page count returned by the upstream provider. Example: {} |
dataSnapshotMonth | string | optional | Data snapshot month returned by the upstream provider. Example: {} |
Examples
Example arguments
{
"type": "product",
"imageUrlList": [
"https://example.com/product.jpg"
],
"keyword": "phone case",
"provider": "default",
"refVideoUrl": "https://example.com/reference-video.mp4",
"saleCountry": "US",
"targetLanguage": "en",
"duration": 15
}Example structuredContent
{
"taskId": "6b3f1c61-1b7d-4a59-8a7d-3b23e4b0c9e2",
"status": "PENDING",
"message": "The task has been created. Use taskId to query the result."
}