Amazon listing image and video MCP server (Amazon Agent Iris)

io.github.KuudoAI由 GitHub 帳號 KuudoAI 發布v1.0.0更新於 Oct 24, 2026

MCP server that creates Amazon listing images and product video from live catalog data.

已驗證Streamable HTTP可網頁執行商業與電商媒體與設計

概覽

AI 產生的概覽

遠端 MCP 伺服器，讓助理在自有 Kuudo 部署中依即時商品目錄資料產生與編輯 Amazon 商品圖片和產品影片。

可以做什麼
使用 Google Gemini 或 OpenAI 模型產生或編輯商品圖片，支援鏈式精修以及圖示、圖樣、圖表、分鏡、照片修復等操作預設，並依從技能資源讀取的 Amazon 圖片合規規則檢查。建立有時效的簽章上傳網址，以便把本機圖片位元組送入產生流程。使用 Google Veo 從文字、首尾影格或參考圖產生產品影片，並查詢、列出與續接長時間執行的影片工作。讀取伺服器的資源與提示詞，包括 Amazon 圖片與影片技能文件。查看伺服器在驗證、模型、儲存、上傳與健康狀態的診斷資訊。
適用情境
適合想讓助理依平台規則製作或稽核商品圖片與產品影片，並執行可續接的影片算圖工作的 Amazon 賣家與商品團隊，且需在自有 Kuudo 雲端部署中使用。
執行需求
需要 Kuudo 部署主機，透過 streamable HTTP 連線到該部署的 MCP 網址。需要 Authorization bearer 標頭，攜帶 Kuudo API 金鑰或部署 bearer 權杖。OpenAI 圖片工具另需 OpenAI 金鑰，同樣以 Authorization bearer 標頭傳送。此儲存庫本身只是僅含目錄的樁程式，實際執行只在部署中進行。
安裝前請注意
Authorization 標頭攜帶 Kuudo API 金鑰或部署 bearer 權杖，OpenAI 圖片工具還要求在同一標頭名稱中提供另一個供應商金鑰。圖片與影片產生由底層供應商計費；若影片提交結果不明，不要用新的 id 重試，因為 Google 可能已接受並計費。寫入工具只在你的部署內運作，README 稱其不會變更 Amazon 上的任何內容。上傳網址為簽章網址且會過期。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Amazon listing image and video MCP server (Amazon Agent Iris)，將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

其他 MCP 客戶端

參照 儲存庫 中的啟動說明。

README

Amazon listing image and video MCP server (Amazon Agent Iris) by Kuudo

Create Amazon listing images and product video from live catalog data, real product references, Amazon rules, and secure, human-approved workflows.

[MCP] [Registry] [License: MIT] [Community]

Product page: https://www.kuudo.com/features/amazon-agent-iris/ · Docs: https://www.kuudo.com/docs/amazon-agent-iris/ · Pricing: https://www.kuudo.com/pricing.md

Connect

Kuudo runs in your own cloud. The Community plan deploys one instance of each Amazon MCP server into your account, and your client connects to that deployment:

json
{  "mcpServers": {    "amazon-agent-iris-mcp": {      "url": "https://<your-host>/mcp",      "headers": {        "Authorization": "Bearer <your Kuudo API key>"      }    }  }}

Replace <your-host> with the hostname of your deployment and the bearer value with your Kuudo API key.

A tool may document a provider credential of its own. That is separate from the bearer above, it belongs to the service the tool calls, and the tool's own entry says so.

What this repository is

This repository holds registry metadata and a catalog-only stub. Live execution runs in your Kuudo deployment. The server source is not published. The stub in src/ answers tools/list with the catalog below, serves the same catalog as one resource (kuudo://catalog/tools.json), offers one prompt (connect) carrying the setup guidance, and returns an error with that guidance on any call, so registries and clients can inspect the surface without any access to Amazon.

Inspect the catalog locally with Docker

The image runs the same catalog-only stub over stdio. It is not the live server.

bash
docker build -t amazon-agent-iris-mcp .docker run -i --rm amazon-agent-iris-mcp

Point a client at it with a stdio entry:

json
{  "mcpServers": {    "amazon-agent-iris-mcp-catalog": {      "command": "docker",      "args": ["run", "-i", "--rm", "amazon-agent-iris-mcp"]    }  }}

Tools

12 tools, read from a running instance of this server. Four of the twelve mirror the MCP resource and prompt methods, and two more are onboarding and server health. The remaining six generate images and video, hand back an upload URL for local source bytes, and track video jobs.

Each tool carries the argument schema the live server publishes, so a client can inspect the full call signature here before it connects to your deployment.

Every tool (12)
ToolAccessWhat it does
start_herereadCall this FIRST to understand how to use this image and video generation server. Returns workflow documentation for: - Available tools - HTTP image uploads - Editing photos - Long-running, resumable video generation - Generating new images - Downloading results
generate_imagewriteGenerate or edit images with Google Gemini (Nano Banana) from natural language instructions. For OpenAI gpt-image-2, use the sibling generate_openai_image tool. For Amazon / Seller Central listing images, FIRST read the skill://amazon-product-image/SKILL.md resource (via read_resource) for the marketplace image-compliance rules, then generate/audit against them. ## CHARACTER CONSISTENCY Use public image handles for character-consistent image generation: generate_image( input_images=["agent-iris://images/img_12"], mode="generate", prompt="PRESERVE EXACTLY: character's facial features. NEW: Professional headshot...", model="gemini-3-pro-image" ) Explicit generate mode uses the image handle as conditioning rather than an edit source. ## EDITING AN EXISTING IMAGE For local bytes, upload with POST /images first, then pass the returned handle. For a public hosted image, pass the HTTPS URL directly: generate_image( input_images=["https://example.com/source.jpg"], mode="edit", prompt="Change the background to a beach sunset" ) ## ITERATING ON A RESULT (high-fidelity chained edit) — PREFERRED Every generate/edit response returns an interaction_id. Pass it back to refine the SAME image while preserving the rest of the composition — no need to re-send the source. This beats re-editing from bytes for edit fidelity: generate_image( interaction_id="v1_Chd...", # from the previous response prompt="Recolor the mug to beige. Change nothing else." ) TTL: ~55 days (paid) / 1 day (free). ## COMPOSITING (Person from image A into scene B) Use multiple references with generate mode: generate_image( input_images=["agent-iris://images/img_12", "gemini://files/beachRef"], mode="generate", prompt="Place character from image 1 into the beach scene from image 2..." ) ## OPERATION PRESETS Use operation="icon", "pattern", "diagram", "storyboard", or "photo_repair" for strict defaults around model, aspect ratio, and prompt framing. ## Mode Auto-Detection: - interaction_id → EDIT mode - operation="photo_repair" → EDIT mode - operation="icon"/"pattern"/"diagram"/"storyboard" → GENERATE mode - operation="general" with one input image → EDIT mode - Multiple input images or pure prompt → GENERATE mode Returns both MCP image content blocks and structured JSON with metadata.
generate_openai_imagewriteGenerate images with OpenAI gpt-image-2 via the Responses API. The additive OpenAI provider, sibling to generate_image (Google Gemini). Send your OpenAI key as Authorization: Bearer <key> — the tool name implies the provider. Iterate on a result with previous_response_id (OpenAI's analog of Gemini's interaction_id).
create_upload_urlwriteCreate the expiring signed URL needed to upload LOCAL image bytes. Use this when you have an image on disk (or raw bytes) to feed into generate_image: call this tool, then HTTP POST the bytes to the returned upload_url (no auth header needed — the URL's signed token authorizes it). The upload responds with a agent-iris://images/{id} handle to pass in generate_image's input_images. Hosted http(s) image URLs do NOT need this — pass them to generate_image directly.
server_statusreadReturn server diagnostics for auth, models, storage, uploads, resources, and health.
generate_videowriteGenerate video with Google Veo 3.1 from text, a first image, first/last frames, up to three reference images, or a Veo-generated video to extend. Inputs are bounded handles or http(s) URLs, never inline data; references, extensions, and 1080p/4k output require an 8-second request. Before planning or generating video, read the canonical skill://amazon-video/SKILL.md resource. It covers mode and model selection, prompt craft, recovery, and the additional compliance workflow for Amazon listing videos. Renders take 11 s to several minutes. Task-capable clients run this in the background; other clients receive pending and resume with get_video_status(recovery_id). Reuse the same recovery_id after a timeout. If submission is reported as unknown, do not retry under a new id because Google may already have accepted the billed request. Results are signed download URLs and agent-iris://videos/vid_N handles.
get_video_statusreadProbe one video operation (exactly one of recovery_id / operation_name). Makes at most one Google status call and never waits for rendering. When the render is finished it downloads and stores the video in this call. Owner-scoped: only the key that submitted the operation can see it.
list_video_operationsreadList the video operations submitted with this key, newest first. Reads only the local operation store — no Google call. Use a listed recovery_id with get_video_status to resume or collect.
list_resourcesreadList all available resources and resource templates. Returns JSON with resource metadata. Static resources have a 'uri' field, while templates have a 'uri_template' field with placeholders like {name}.
read_resourcereadRead a resource by its URI. For static resources, provide the exact URI. For templated resources, provide the URI with template parameters filled in. Returns the resource content as a string. Binary content is base64-encoded.
list_promptsreadList all available prompts. Returns JSON with prompt metadata including name, description, and optional arguments.
get_promptreadGet a prompt by name with optional arguments. Returns the rendered prompt as JSON with a messages array. Arguments should be provided as a dict mapping argument names to values.

Set up your client

Playbooks

Operator guides grounded in Amazon's own documentation, each with the artifact the agent produces:

How it compares

Reads, writes, approvals

Read tools are safe to call freely. Write tools do work inside your deployment and never change anything on Amazon. Each tool's access is recorded beside it in tools.json.

Kuudo

Generated from KuudoAI/marketing. Do not edit by hand; changes are overwritten on the next sync.

來源：README.md，提交 3dafb84

工具

0
SourceWeft 尚未讀取此伺服器的工具。

版本歷史

1
  1. v1.0.0最新Sep 16, 2026

評分

已滿足適用檢查項的 60%

該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分

    請保持端點能夠回應 MCP 握手；SourceWeft 每 30 天探測一次。

  • 已宣告認證方式必需10 分
  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list，或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分