Google Cloud Vision
Google Cloud Vision API enables developers to integrate vision detection features into applications, including image labeling, face and landmark detection, optical character recognition (OCR), and explicit content tagging.
Verdict
Common use cases
- Extract invoice text from email attachments
- Flag user-uploaded photos for explicit content
- Identify landmarks in travel photos for tagging
- Build visual product search from catalog images
- Audit brand assets for logo placement
Integration
- Vendor
- Google Cloud Vision
- Category
- other
- Auth
- API_KEY
- Tools
- 23
- Composio slug
google_cloud_vision
Tools
- Add Product to ProductSet
Tool to add a Product to a specified ProductSet. Use after creating both resources in the same project/location to link a product to its set.
- Cancel Vision Operation
Tool to cancel a long-running Vision API operation. Use when you need to abort a pending or in-progress operation.
- Create ReferenceImage
Tool to create a ReferenceImage under a product. Use when adding a new image to a product for detection.
- Create Vision Product
Tool to create and return a new Product resource. Use when you need to register a product in a specific project/location after preparing product details.
- Delete Productdestructive
Tool to permanently delete a Product and its reference images. Use after confirming the product's resource name.
- Delete Product Setdestructive
Tool to permanently delete a ProductSet. Use after confirming the ProductSet's resource name.
- Delete Reference Imagedestructive
Tool to permanently delete a reference image. Use when you have confirmed the reference image's resource name.
- Delete Vision API Operationdestructive
Tool to delete a long-running Vision API operation. Use after confirming the operation name.
- Get Product
Tool to get information associated with a Product. Use when you have the product resource name and need its details.
- Get Product Set
Tool to get a ProductSet. Use when you need metadata details of an existing ProductSet by its full resource name. Use after obtaining the resource name.
- Get Reference Image
Tool to get information associated with a ReferenceImage. Use when you have the full resource name and need its metadata.
- Get Vision API Operation
Tool to get the latest state of a long-running operation. Use after starting an async Vision API operation to poll its status.
- Import Product Sets
Tool to asynchronously import reference images into ProductSets from a CSV in GCS. Use when you need to bulk import images into product sets via a Cloud Storage CSV.
- List IndexEndpoints
Tool to list IndexEndpoints in a project and location. Use when you need to retrieve existing IndexEndpoints and handle pagination.
- List Locations
Tool to list available Vision AI service locations for a project. Use when you need to discover supported regions before making region-specific API calls.
- List Products in ProductSet
Tool to list Products in a specified ProductSet. Use when you need to retrieve Products associated with a ProductSet after confirming it exists, with optional pagination.
- List Projects
Tool to list Google Cloud projects accessible by the authenticated user. Use when you need to enumerate available project IDs and resource names before performing further operations.
- List Reference Images
Tool to list reference images for a product. Use when you need to retrieve stored reference images under a specified product resource name, with optional pagination.
- List Vision API Operations
Tool to list operations that match the specified filter. Use when you need to retrieve all operations under a specific project and location.
- Purge Productsdestructive
Tool to asynchronously delete products in a ProductSet or orphan products. Use when you need to clean up products at scale; ensure `force` is true to execute.
- Remove Product from ProductSetdestructive
Tool to remove a Product from a specified ProductSet. Use after creating both resources in the same project/location to unlink a product from its set.
- Update Product
Tool to update a Product's mutable fields: displayName, description, and productLabels. Use after confirming the product resource name.
- Update Product Set
Tool to update a ProductSet resource. Use when you need to modify the displayName of an existing ProductSet.
Setup
Setup guide
- 11. Open your Google Cloud Console and enable the Vision API for your project. 2. Navigate to APIs & Services > Credentials and create a new API key (restrict it to Vision API for security). 3. In Switchy, click Integrations in the left sidebar, find Google Cloud Vision, and click Connect. 4. Paste your API key when prompted and click Authorize. 5. Open any Space and type '@Google Cloud Vision analyze this image' followed by an image URL or attachment to test the connection. 6. If the MCP returns labels, faces, or text detected in the image, the connection works. 7. For product catalog tools (add products, create reference images), ensure your API key's associated service account has roles/vision.productSearchEditor role in IAM. 8. To invoke specific tools, @mention the MCP and describe the task — it will route to the appropriate detection or catalog management endpoint.
What teammates see: by default, memories from Google Cloud Vision are scoped to the Space (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Spaces have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Analyze Support Screenshot
@Google Cloud Vision detect all text in this screenshot and return it as plain text so I can search for error codesOpen in a Space →
Moderate User Upload
@Google Cloud Vision analyze this image for adult, violent, or racy content and return likelihood scores for each categoryOpen in a Space →
Tag Product Photo
@Google Cloud Vision detect all objects and labels in this product image and list them with confidence scores above 80%Open in a Space →
Identify Landmark
@Google Cloud Vision identify any landmarks in this photo and return their names and geographic coordinatesOpen in a Space →
Add Product to Catalog
@Google Cloud Vision create a new product named 'Blue Canvas Sneaker' in project my-retail-project location us-west1 and add it to ProductSet shoes-spring-2024Open in a Space →
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@Google Cloud Vision analyze this product image and tell me what objects you detect, including their confidence scores
Based on the image analysis, I detected the following objects: a ceramic coffee mug (confidence: 94%), a wooden table surface (confidence: 89%), and a smartphone (confidence: 87%). The mug appears to be white with a handle on the right side. The detection also identified text on the mug but would require a separate OCR request to read it. The image dimensions analyzed were 1920x1080 pixels with good lighting conditions, which contributed to the high confidence scores.
This example showcases Vision API's core object detection capability. Requires an API key with Vision API enabled and sufficient quota. Note that detection quality depends on image resolution and lighting—low-quality images may return lower confidence scores or miss objects entirely.
@Google Cloud Vision create a new product called 'Vintage Leather Wallet' in the us-west1 location under project acme-retail, then add it to the ProductSet 'accessories-2024'
I've successfully created the product 'Vintage Leather Wallet' with resource name projects/acme-retail/locations/us-west1/products/prod_abc123. The product has been added to ProductSet 'accessories-2024'. You can now add reference images to this product using the Create ReferenceImage tool. The product is now indexed and will be available for visual search queries within 5-10 minutes as the Vision API processes the ProductSet updates.
This demonstrates the product management workflow for Vision Product Search. Requires write permissions and a valid project with billing enabled. The ProductSet must exist before adding products to it. Be aware that product indexing isn't instantaneous—allow processing time before running search queries against newly added products.
@Google Cloud Vision I have 15 product images in my staging bucket. Create products for each, add reference images, and summarize which ones had the highest quality scores for visual search
I've processed all 15 images from your staging bucket and created corresponding products. Quality analysis summary: 8 images scored 'excellent' (>90% quality) with clear backgrounds and good lighting, 5 scored 'good' (75-90%) with minor occlusion or angle issues, and 2 scored 'fair' (<75%) due to blur or poor contrast. The highest-scoring image was 'leather-boot-01.jpg' at 96%, ideal for visual search. The two fair-quality images (handbag-03.jpg, wallet-08.jpg) may need retakes for better search performance. All products are now in ProductSet 'staging-review' awaiting your approval.
This example combines multiple Vision API operations with AI reasoning to provide actionable insights. Requires API quota for batch operations (15+ API calls) and Cloud Storage read permissions. Processing time scales with image count and size—expect 30-60 seconds for this batch. Quality scores help prioritize which product photos need improvement before going live.
Use-case deep-dives
When Vision API makes sense for product catalog workflows
A 6-person e-commerce team uploads 200-300 product photos weekly and needs consistent tagging for search and merchandising. Google Cloud Vision's product-set tools (Create Vision Product, Add Product to ProductSet, Create ReferenceImage) let you build a reference library that auto-tags new uploads against your existing catalog. The API_KEY auth keeps setup simple—no OAuth dance for batch jobs. The trade-off: if your catalog is under 500 SKUs or you're just doing one-off image analysis, the 23-tool surface area is overkill. Vision shines when you're managing a growing catalog where new images need to match existing products. If your team runs weekly catalog updates and wants to catch duplicate listings or suggest similar items, this MCP pays off in the first sprint.
Why Vision API isn't the first choice for moderation workflows
A 3-person community platform reviews user-uploaded images for policy violations before publishing. Google Cloud Vision can detect explicit content and label objects, but this MCP exposes the product-search tooling (ProductSets, ReferenceImages) instead of the core detection APIs your moderation flow actually needs. You'll spend more time working around the product-focused tools than building your review queue. The 23-tool count suggests breadth, but most are CRUD operations for managing product catalogs—not the label-detection or safe-search calls a moderation pipeline wants. If your use case is flagging inappropriate uploads in real time, look for an MCP that surfaces Vision's detection endpoints directly. This one is built for catalog management, not content safety workflows.
When Vision's reference-image system fits design-team asset lookup
A 5-person design team maintains 2,000+ brand assets (logos, product shots, UI components) and wastes 30 minutes daily hunting for the right file. Google Cloud Vision's ReferenceImage and ProductSet tools let you index your asset library so designers can upload a rough screenshot and find the canonical version. The API_KEY auth means your ops person can set it up without involving IT for service-account provisioning. The catch: if your library is static or under 500 assets, a simple file-naming convention beats the overhead of managing ProductSets. Vision wins when your asset count grows weekly and you need fuzzy matching—designers searching by visual similarity, not exact filenames. If your team spends more time searching than creating, this MCP turns that 30-minute hunt into a 30-second lookup.
Frequently asked
What does the Google Cloud Vision MCP do in Switchy?
It connects Switchy to Google Cloud Vision's image analysis APIs. Your team can create and manage product catalogs, add reference images, and run visual search operations without writing code. The MCP handles 23 Vision API operations including product sets, reference images, and long-running operation management. You still need a Google Cloud project with Vision API enabled.
Do I need a Google Cloud account to use this MCP?
Yes. You need an active Google Cloud project with the Vision API enabled and billing configured. The MCP authenticates via API key, so you'll generate one in your GCP console with Vision API permissions. Without a project and valid key, the integration won't connect. This isn't a free trial—Google charges for Vision API usage per your GCP billing.
Can this MCP analyze images or just manage product catalogs?
This MCP focuses on product catalog management—creating products, adding reference images, organizing ProductSets, and canceling operations. It doesn't expose Vision's core image analysis features like label detection, OCR, or face detection. For those, you'd call Vision's REST API directly or use a different integration. Think of this as the catalog admin layer, not the image intelligence layer.
How is this different from calling the Vision API directly?
The MCP wraps Vision's catalog management endpoints in natural language tools your team can use in Switchy without reading API docs. You skip writing HTTP requests, handling auth headers, and parsing JSON responses. The trade-off: you're limited to the 23 operations the MCP exposes. For custom workflows or newer Vision features, the REST API gives you more control.
Who on my team should connect this integration?
Whoever manages your Google Cloud project and can generate API keys. That person needs IAM permissions to enable Vision API and create credentials. Once connected in Switchy, any team member with workspace access can use the tools—but they'll be operating under the same GCP project and billing account. Don't share the API key outside Switchy.