Sirv AI Studio

io.github.sirv由 GitHub 帳號 sirv 發布v1.1.1更新於 Nov 4, 2026

AI image, video and 3D tools, assets, products and Shopify for Sirv AI Studio.

已驗證Streamable HTTP可網頁執行檔案與儲存商業與電商媒體與設計

概覽

AI 產生的概覽

Sirv AI Studio 伺服器,用於 AI 影像、影片與 3D 處理,並管理素材、商品、Shopify 與 Sirv CDN 檔案。

可以做什麼
  • 處理影像:去除或替換背景、放大、移除物件、產生深度圖、產生替代文字
  • 生成與轉換媒體:文字生圖、圖生圖、圖片內文字翻譯、影片生成與放大、圖片轉 3D 及 GLB 最佳化
  • 製作商品視覺:生活情境商品照、虛擬試穿、批次去背與批次放大
  • 管理素材與商品:列出、搜尋、尋找相似、更新、刪除、加標籤、依網址匯入,並將素材連結到商品
  • 串接 Shopify 與供應商入口:檢查連線、同步商品、推送圖片、查看提交需求並執行自動修正
  • 管理 Sirv CDN 檔案:搜尋、讀取與更新中繼資料和標籤、建立資料夾、依網址上傳、重新命名、複製或刪除
適用情境
適合希望由助理產生或修整商品影像、影片與 3D 模型,並在 Sirv AI Studio 工作區中整理這些成果(包含推送到 Shopify)的情境。也適合已在使用 Sirv 素材、商品或 CDN 檔案、想透過對話完成這些操作的團隊。
執行需求
可使用託管端點,或在本地以 Node.js 執行 npm 套件(npx)。驗證方式為 OAuth,或透過 SIRV_AI_API_KEY 提供 API 金鑰;沒有金鑰時伺服器會透過瀏覽器登入。需要具備額度的 Sirv AI Studio 帳號,且影像輸入必須是可公開存取的網址。
安裝前請注意
多數 AI 工具會消耗 Studio 額度,圖片轉 3D 與影片生成費用較高,供應商入口自動修正明確按額度計費。許多工具會寫入、覆寫或刪除資料,包括批次刪除以及 Shopify 發佈與同步。影像會經 fal.ai 的 API 處理並受其資料政策約束,OAuth 權杖保存在本機權杖檔案中。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Sirv AI Studio,將其新增到工作區。
  2. 連接 所需的 憑證。它們會被私密儲存,絕不會在公開頁面上展示。
  3. 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。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式,或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南,先向你展示發布者、端點和登入方式,經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.sirv%2Fsirv-ai-mcp-server/install.md,並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前,先告訴我它的發布者、端點和登入方式,等我確認。不要把 API 金鑰寫進命令或我會提交的檔案;問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。 客戶端會在首次連接時完成 OAuth 登入。

{
  "mcpServers": {
    "sirv-ai-studio": {
      "type": "http",
      "url": "https://mcp.sirv.studio/"
    }
  }
}

README

Sirv AI MCP Server

MCP (Model Context Protocol) server for Sirv AI Studio - AI-powered image and video processing.

Features

This MCP server exposes 54 tools across AI processing, asset management, product management, Shopify integration, Supplier Portal operations and Sirv CDN file management:

Image Processing

  • sirv_remove_background - Remove backgrounds from images using BiRefNet or Bria
  • sirv_upscale - Upscale images 2-4x using ESRGAN, Clarity, or Topaz
  • sirv_background_replace - Replace image backgrounds with AI-generated scenes
  • sirv_object_removal - Remove objects from images, or fill a masked area from a prompt
  • sirv_depth_map - Generate depth maps from images

Image Generation

  • sirv_generate - Generate images from text prompts
  • sirv_image_to_image - Transform images based on text prompts
  • sirv_image_translation - Translate text inside images

Product AI

  • sirv_product_lifestyle - Create lifestyle product shots
  • sirv_virtual_try_on - Virtual clothing try-on

3D & Video

  • sirv_image_to_3d - Convert images to 3D models
  • sirv_optimize_glb - Optimize GLB models with mesh and texture compression
  • sirv_video_generation - Generate videos from text/images
  • sirv_video_upscale - Upscale videos 2x or 4x to listing quality with SeedVR or Topaz

Batch Operations

  • sirv_batch_remove_background - Process entitled batches of images
  • sirv_batch_upscale - Upscale entitled batches of images

Workflows

  • sirv_list_workflows - List saved workflows
  • sirv_get_workflow - Get workflow details

Asset Management (DAM)

  • sirv_list_assets - List assets with filtering and pagination
  • sirv_get_asset - Get a single asset by ID
  • sirv_search_assets - Search workspace assets by text, matching every word or its stem
  • sirv_find_similar_assets - Find visually similar assets using the workspace visual-similarity index
  • sirv_update_asset - Update asset fields (filename, alt text, favorite)
  • sirv_delete_asset - Soft-delete an asset
  • sirv_get_asset_metadata - Get custom metadata for an asset
  • sirv_update_asset_metadata - Update custom metadata fields
  • sirv_generate_asset_alt_text - AI-generate alt text for an asset
  • sirv_bulk_asset_operation - Bulk favorite/unfavorite/delete/tag/move
  • sirv_import_asset_url - Import an asset from an external URL

Product Management (PIM)

  • sirv_list_products - List products with filtering and pagination
  • sirv_get_product - Get a single product by ID
  • sirv_create_product - Create a new product
  • sirv_update_product - Update product fields
  • sirv_delete_product - Delete a product
  • sirv_search_products - Quick search by name or SKU
  • sirv_list_products_with_assets - List products with linked asset details
  • sirv_link_assets_to_product - Link assets to a product
  • sirv_bulk_product_action - Bulk delete, update status, link assets

Shopify Integration

  • sirv_shopify_status - Check Shopify connection status
  • sirv_push_image_to_shopify - Publish a Studio asset to a Shopify product through an approved publish plan
  • sirv_sync_shopify_products - Start a background import of Shopify products and images into the PIM

Supplier Portal

  • sirv_get_supplier_sftp_status - Check entitled Supplier Portal SFTP connection details
  • sirv_get_supplier_review_assist_requirements - Check Supplier Portal submission requirements
  • sirv_run_supplier_autofix - Run credit-charged autofix on a supplier image

Sirv CDN File Management

These tools work on the workspace's connected Sirv account, rather than the Studio asset library:

  • sirv_cdn_search_files - Search Sirv files with pagination and sorting
  • sirv_cdn_get_file_metadata - Read file metadata
  • sirv_cdn_manage_folder - Create a folder or list its contents
  • sirv_cdn_upload_from_url - Upload a file from a public URL
  • sirv_cdn_update_file_metadata - Set title, description, tags or product metadata
  • sirv_cdn_update_file_tags - Add or remove file tags
  • sirv_cdn_manage_file - Rename, copy or delete a file

Utility

  • sirv_alt_text - Generate alt text descriptions
  • sirv_get_usage - Check credit usage statistics

Organization Context

  • sirv_list_organizations - List the organizations available to the authenticated user

All DAM, product, Shopify, Supplier Portal and Sirv CDN tools accept an optional orgId parameter to scope requests to a specific organization. If omitted, the user's default organization is used. The server sends orgId as the X-Org-Id header, which the API validates against the user's org membership.

Entitlement Checks

Every Studio plan includes API and MCP access; an admin can turn it off for a workspace. Preview-readable tools such as list, get, search, metadata-read, usage, organization, and workflow-read calls then execute without another tool-specific check. Mutation tools, direct AI tools that spend credits, batch operations, visual similarity, Shopify delivery/sync, and Supplier Portal Pro tools call Studio's /api/mcp/authorize-tool endpoint before execution and fail closed when the workspace lacks the matching entitlement.

Prompts

  • prepare-listing-image (assetId) - Remove the background, upscale if small, and write alt text for one library image
  • alt-text-sweep (folderId) - Find the images in a folder that have no alt text and write it, after the user confirms the cost

Installation

Use the hosted server at https://mcp.sirv.studio, or run the npm package:

bash
npx -y sirv-ai-mcp-server

Use version 1.0.5 or later. Earlier versions are the legacy server from before the www.sirv.studio cutover on 2026-10-30.

To build from source:

bash
bun installbun run buildnode dist/index.js

Configuration

Environment Variables

bash
# Optional: Sirv AI Studio URL (default: https://www.sirv.studio)SIRV_AI_BASE_URL="https://www.sirv.studio"
# Optional: API key for authentication (if not using OAuth)SIRV_AI_API_KEY="your-api-key"
# Transport mode: "stdio" (default) or "http"MCP_TRANSPORT="stdio"
# HTTP server settings (when using http transport)MCP_HTTP_PORT=3001MCP_HTTP_HOST=0.0.0.0MCP_BASE_URL="https://your-mcp-server.com"
# HTTPS (optional - can also use reverse proxy)MCP_SSL_KEY_PATH="/path/to/key.pem"MCP_SSL_CERT_PATH="/path/to/cert.pem"
# Security: IP allowlistingMCP_ENABLE_IP_ALLOWLIST="true"MCP_ALLOWED_IPS="1.2.3.4,5.6.7.8/24"
# CORS: Allowed origins (comma-separated)MCP_CORS_ORIGINS="https://claude.ai,https://your-app.com"

Transport Modes

Stdio Transport (Default)

Used for local integrations with Claude Desktop and Claude Code.

bash
bun run start# orbun run dev

HTTP Transport

Used for remote/hosted deployments accessible via HTTPS.

bash
bun run start:http# orbun run dev:http

When running in HTTP mode, the server exposes:

  • / - MCP protocol endpoint (/mcp is also accepted for compatibility)
  • /health - Health check endpoint
  • /.well-known/oauth-authorization-server - OAuth 2.0 metadata
  • /.well-known/oauth-protected-resource - Protected resource metadata

Security Features

Request Limits

HTTP requests accept uncompressed JSON up to 1 MiB. Tool arguments are limited to 10,000 array elements and object members in total, including nested values. Split larger requests into smaller calls. File uploads use URLs rather than inline bytes.

Each signed-in caller can send 100 requests a minute. One IP address can send 30 invalid API keys a minute. Expired OAuth tokens do not count toward that limit.

OAuth 2.0 Support

The server implements OAuth 2.0 with PKCE for secure authentication:

  • Authorization code flow with S256 code challenge
  • Token refresh support
  • Secure token storage in ~/.sirv-ai/tokens.json

Tool Safety Annotations

All tools include proper MCP safety annotations:

  • readOnlyHint - Indicates if the tool only reads data
  • destructiveHint - Indicates if the tool can delete or overwrite data
  • idempotentHint - Indicates if repeated calls produce the same result
  • openWorldHint - Indicates if the tool accesses external services

CORS Configuration

When running in HTTP mode, CORS is configured for browser-based authentication:

  • Default allowed origins: https://claude.ai, https://console.anthropic.com
  • Customizable via MCP_CORS_ORIGINS environment variable
  • Proper preflight handling with credentials support

IP Allowlisting

Optional IP-based access control:

  • Pre-configured Claude.ai/Claude Code IP ranges
  • Enable with MCP_ENABLE_IP_ALLOWLIST=true
  • Add custom IPs/CIDR ranges via MCP_ALLOWED_IPS

HTTPS Support

For production deployments:

  1. Direct HTTPS: Set MCP_SSL_KEY_PATH and MCP_SSL_CERT_PATH
  2. Reverse Proxy: Run HTTP behind nginx/Caddy with TLS termination

Usage with Claude Desktop

Add to your Claude Desktop configuration (claude_desktop_config.json):

json
{  "mcpServers": {    "sirv-ai": {      "command": "npx",      "args": ["-y", "sirv-ai-mcp-server"],      "env": {        "SIRV_AI_BASE_URL": "https://www.sirv.studio"      }    }  }}

Usage with Claude Code

Add to your Claude Code MCP settings:

json
{  "mcpServers": {    "sirv-ai": {      "command": "npx",      "args": ["-y", "sirv-ai-mcp-server"],      "env": {        "SIRV_AI_BASE_URL": "https://www.sirv.studio"      }    }  }}

Usage with ChatGPT

For the hosted server, use https://mcp.sirv.studio as the MCP connector URL. Connectors added with https://mcp-dev.sirv.studio keep working. Studio handles sign-in and consent at https://www.sirv.studio. Every Studio plan can connect. When a workspace runs out of credits, tools answer with a link to Studio's Buy credits page. This connection path is separate from ChatGPT Actions, which use the OpenAPI specification.

Use the callback URL supplied by your ChatGPT connector. Studio accepts https://chatgpt.com/connector/oauth/<id> with one ID segment containing letters, digits, underscores or hyphens, as well as the older https://chatgpt.com/connector_platform_oauth_redirect callback. The URL must also match the client's registered redirect URI. OAuth still requires PKCE S256. See the MCP developer docs.

In the current source, ten image-result tools attach an interactive image widget. Account, asset, product, batch, workflow, video and 3D tools return their tool results without that widget. For multi-image generation, the widget previews the first result rather than a gallery.

The widget uses the MCP Apps resource ui://widget/sirv-studio-v3.html with MIME type text/html;profile=mcp-app. This describes the server implementation, not a claim that the app is listed in the ChatGPT directory.

Remote Server Deployment

For hosting the MCP server remotely (accessible via claude.ai):

Keep the two origins distinct:

  • SIRV_AI_BASE_URL is the Studio app that serves OAuth and API requests.
  • MCP_BASE_URL is the public origin clients use to reach the MCP server. It supplies the advertised OAuth resource and widget domain.

The Docker Compose configuration defaults to https://www.sirv.studio and https://mcp.sirv.studio, respectively. Set both explicitly for another deployment.

Check discovery before connecting a client:

bash
curl --fail https://mcp.sirv.studio/.well-known/oauth-protected-resourcecurl --fail https://mcp.sirv.studio/.well-known/oauth-authorization-server

For these defaults, resource should be https://mcp.sirv.studio, and issuer and the sole authorization_servers entry should be https://www.sirv.studio. The authorization, token and registration endpoints should use the Studio origin.

1. Deploy with HTTPS

bash
# Using SSL certificates directlyMCP_TRANSPORT=http \MCP_HTTP_PORT=443 \MCP_BASE_URL=https://mcp.your-domain.com \MCP_SSL_KEY_PATH=/etc/ssl/private/key.pem \MCP_SSL_CERT_PATH=/etc/ssl/certs/cert.pem \MCP_ENABLE_IP_ALLOWLIST=true \SIRV_AI_BASE_URL=https://www.sirv.studio \bun run start:http

2. Or use a reverse proxy (recommended)

nginx configuration example:

nginx
server {    listen 443 ssl http2;    server_name mcp.your-domain.com;
    ssl_certificate /etc/ssl/certs/cert.pem;    ssl_certificate_key /etc/ssl/private/key.pem;
    location / {        proxy_pass http://127.0.0.1:3001;        proxy_http_version 1.1;        proxy_set_header Upgrade $http_upgrade;        proxy_set_header Connection 'upgrade';        proxy_set_header Host $host;        proxy_set_header X-Real-IP $remote_addr;        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;        proxy_set_header X-Forwarded-Proto $scheme;    }}

3. Configure in claude.ai

Once deployed, users can connect via the MCP settings in claude.ai using your server URL.

Development

bash
# Run in development mode with auto-reload (stdio)bun run dev
# Run in development mode (http)bun run dev:http
# Build for productionbun run build
# Run the server (stdio)bun run start
# Run the server (http)bun run start:http

Example Tool Usage

Remove Background

Remove the background from https://example.com/product.jpg

Generate Image

Generate an image of a modern minimalist living room with natural lighting

Browse Assets

Show me my latest 10 assetsSearch my assets for "product photos"

Manage Products

List all active products from ShopifyCreate a product with SKU "CHAIR-001" named "Modern Chair"Link these 3 assets to the chair product as gallery images

Shopify Workflow

Check my Shopify connection statusSync products from Shopify including imagesPush this processed image to Shopify product 12345

Credit Costs

ToolCredits
Background Removal1-2
Upscale2-3
Generate2
Image to Image2-4
Background Replace4
Object Removal3
Alt Text1
Depth MapFREE
Virtual Try-On4
Image to 3D80
Video GenerationVariable

Troubleshooting

Connection Issues

"Server not found" or connection refused

  • Verify the server is running: bun run start for stdio or bun run start:http for HTTP mode
  • Check that SIRV_AI_BASE_URL is set correctly
  • For HTTP mode, ensure the port isn't blocked by a firewall

"Authentication failed"

  • Delete stored tokens: rm -rf ~/.sirv-ai/tokens.json
  • Re-authenticate through the OAuth flow
  • Verify your account is active at https://www.sirv.studio

"Invalid API key"

  • Check that SIRV_AI_API_KEY is correctly set (if using API key auth)
  • Regenerate your API key in your Sirv AI Studio account settings

Tool Errors

"Insufficient credits"

"Image URL not accessible"

  • Ensure the image URL is publicly accessible (not behind authentication)
  • Try uploading the image to a public hosting service first

"Processing timeout"

  • Large images or complex operations may take longer
  • Try with a smaller image or lower resolution setting
  • Video generation and 3D conversion can take several minutes

HTTP Transport Issues

CORS errors in browser

  • Add your origin to MCP_CORS_ORIGINS environment variable
  • Example: MCP_CORS_ORIGINS="https://your-app.com,https://claude.ai"

SSL/TLS certificate errors

  • Verify certificate paths are correct in MCP_SSL_KEY_PATH and MCP_SSL_CERT_PATH
  • Ensure certificates are not expired
  • For development, consider using a reverse proxy with Let's Encrypt

IP blocked

  • If using IP allowlisting, add your IP to MCP_ALLOWED_IPS
  • Check if your IP has changed (dynamic IP addresses)

Getting Help

Privacy Policy & Terms of Service

By using Sirv AI MCP Server, you agree to our terms and policies:

Data Handling

  • Images are processed through fal.ai APIs and are subject to their data policies
  • Generated outputs are stored temporarily for retrieval
  • User authentication tokens are stored locally in ~/.sirv-ai/tokens.json
  • No user data is collected by the MCP server itself beyond what's necessary for authentication

License

This project is licensed under the MIT License - see the LICENSE file for details.

來源:README.md,提交 b66c2c8

工具

0
該伺服器需要登入,因此 SourceWeft 無法在沒有帳號的情況下讀取其工具,也未收錄。已連接帳號所能看到的內容可能因帳號及其權限而異。

版本歷史

1
  1. v1.1.1最新Nov 4, 2026

評分

已滿足適用檢查項的 85%

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
  • 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