Google Cloud Vision for AI agents
Google Cloud Vision analyzes images for labels, faces, landmarks, text, and explicit content, giving apps computer-vision capabilities via API. Teams building AI products use it to add this capability without training or hosting their own model. Connect it once through Arc0, and your agent, or Claude, ChatGPT and Cursor, can use it through one MCP endpoint, limited to what each user approved.
What agents do in Google Cloud Vision.
Extract text from an image
Run OCR on a scanned receipt or document image to pull out the text for a record.
Label objects in a photo
Get label annotations for an uploaded photo to auto-tag it in a media library.
Check content before publishing
Screen an image for explicit content, read-only, before it is approved for publishing.
29 Google Cloud Vision actions, graded by risk.
Every Google Cloud Vision action is tagged read, write or destructive, so one policy covers the whole app and new actions inherit the right default.
read
5Look things up. Allowed by default.
- google_cloud_vision.get_productGet Product
- google_cloud_vision.list_locationsList Locations
- google_cloud_vision.get_product_setGet Product Set
- google_cloud_vision.list_operationsList Vision API Operations
- google_cloud_vision.list_index_endpointsList Vision AI IndexEndpoints
write
17Create and change things. Allow, or ask the user first.
- google_cloud_vision.create_productCreate Vision Product
- google_cloud_vision.update_productUpdate Product
- google_cloud_vision.create_product_setCreate Product Set
- google_cloud_vision.update_product_setUpdate Product Set
- google_cloud_vision.create_reference_imageCreate ReferenceImage
- google_cloud_vision.annotate_filesAnnotate Files with Vision API
- google_cloud_vision.annotate_imagesAnnotate Images
- google_cloud_vision.import_product_setsImport Product Sets
- google_cloud_vision.vision_get_operationGet Vision API Operation
- google_cloud_vision.vision_list_projectsList Projects
- google_cloud_vision.annotate_location_imagesAnnotate Location Images
- google_cloud_vision.annotate_files_async_batchAsync Batch Annotate Files
- google_cloud_vision.vision_get_reference_imageGet Reference Image
- google_cloud_vision.annotate_images_async_batchAnnotate Images Async Batch
- google_cloud_vision.vision_list_reference_imagesList Reference Images
- google_cloud_vision.vision_add_product_to_product_setAdd Product to ProductSet
destructive
7Delete, cancel or archive. Ask first, or deny outright.
- google_cloud_vision.delete_productDelete Product
- google_cloud_vision.purge_productsPurge Products
- google_cloud_vision.vision_cancel_operationCancel Vision Operation
- google_cloud_vision.vision_delete_operationDelete Vision API Operation
- google_cloud_vision.vision_delete_product_setDelete Product Set
- google_cloud_vision.vision_delete_reference_imageDelete Reference Image
- google_cloud_vision.vision_remove_product_from_product_setRemove Product from ProductSet
Google Cloud Vision in three steps.
- 01Your users connect Google Cloud VisionThey add their Google Cloud Vision api key on Arc0 Connect, under your brand. It goes straight into the vault.
- 02You set the rulesReads run, writes like “create Vision Product” can wait for the user, and “delete Product” can be denied outright.
- 03Any agent can actYour agent calls Google Cloud Vision through the Arc0 SDK or MCP, and so can Claude, ChatGPT and Cursor. Every call lands on the audit log.
await arc0.policies.set('google_cloud_vision', { read: 'allow', write: 'ask', // create_product destructive: 'deny', // delete_product }) # Claude Code: the same connection, one URL $ claude mcp add --transport http arc0 \ https://mcp.arc0.ai/u/u_8f2
How Google Cloud Vision connects.
Users add their Google Cloud Vision api key on Arc0 Connect. It is encrypted in the vault, never shown to the model, and each user can rotate or revoke it at any time.
The same Google Cloud Vision connection serves your agent over MCP and your own backend over REST and the proxy, so a user connects once. How Arc0 handles credentials →
- AUTH
- API key
- CREDENTIALS
- Per-tenant encrypted vault
- MODEL SEES
- Results only, never credentials
- AUDIT LOG
- Every call, on every plan
Use Google Cloud Vision from any agent.
Google Cloud Vision and Arc0, answered.
Can I use Google Cloud Vision with Claude, ChatGPT or Cursor?
Yes. Connect Google Cloud Vision to Arc0 once, then add your Arc0 MCP URL to Claude, ChatGPT, Cursor, Claude Code or any other remote-MCP client. Each assistant only gets the Google Cloud Vision actions you allow.
How do users connect Google Cloud Vision?
Users add their Google Cloud Vision api key on Arc0 Connect. It is encrypted in the vault, never shown to the model, and each user can rotate or revoke it at any time.
Which Google Cloud Vision actions can my agent take?
29 in total: 5 read, 17 write and 7 destructive, such as “create Vision Product”. Your policies decide which of them each agent may call.
Can I stop my agent from deleting things in Google Cloud Vision?
Yes. Actions like “delete Product” are graded destructive. Set destructive actions to deny, or to ask so the user approves each one, and blocked calls still show up on the audit log.
Can my own backend call Google Cloud Vision too?
Yes. The same Google Cloud Vision connection is available over REST and through the Arc0 proxy, so your product and your agent share one connection per user.
Plug Google Cloud Vision into your agent.
Your users connect Google Cloud Vision once, under your brand. Your agent gets 29 actions behind your policies, with every call on the record.
Free to build · MCP + REST · Audit log on every plan