Skip to content
Google Vertex AI logo

Connectors / AI · HTTP Connector

Automate Google Vertex AI with Tray.ai

Generate content, create an endpoint and create a batch prediction job in Google Vertex AI.

What can you do with Google Vertex AI on Tray.ai?

Google Vertex AI is Google Cloud's service for running Gemini and other models, with tuning, batch prediction, and model deployment. Tray.ai connects to Google Vertex AI through the HTTP Connector, so your workflows can call the Google Vertex AI API alongside the rest of your stack.

  • Generate content to draft a reply for each new support ticket, then send it for agent review.
  • Embed content from your knowledge base so your search tool can find answers.
  • Create a batch prediction job each night to score leads exported from your CRM.
Connects through
HTTP Connector
Authentication
OAuth 2.0
Base URL
https://aiplatform.googleapis.com

Set it up

Connect Google Vertex AI to Tray.ai

Each step uses Google Vertex AI's own API details, from its documentation.

Step 1

Sign in to Google Vertex AI with OAuth 2.0

Set up an OAuth 2.0 authentication with the client ID and secret of an OAuth app you register with Google Vertex AI. Sign in once and every step reads the token from there.

Step 2

Point the HTTP Client at the API

Add an HTTP Client step and use https://aiplatform.googleapis.com as the base URL. Or paste a cURL example from Google Vertex AI's API docs, and Tray.ai fills in the method, URL, and headers.

Step 3

Start with “List endpoints”

It's one of 22 documented actions, under Endpoints. The response comes back as data the next step can map into your other apps.

Step 4

Hand it to an AI agent

Wrap “Generate content” and “Stream generated content” in workflows, and Tray.ai's Agent Gateway for MCP publishes them as governed tools Claude and other AI agents can call.

Workflows to build with Google Vertex AI

Each one uses actions and events from Google Vertex AI's own API documentation.

Summarize new Zendesk tickets with Gemini on Vertex AI

When

A ticket is created in Zendesk

Then

  1. Generate content that summarizes the ticket and suggests a category
  2. Add the summary as an internal note on the ticket in Zendesk

With Zendesk

Embed updated Confluence pages for internal search

When

A page is updated in Confluence

Then

  1. Embed content from the page text
  2. Store the embedding with the page link in PostgreSQL

With Confluence and PostgreSQL

Run nightly Vertex AI batch predictions on BigQuery data

When

On a schedule

Then

  1. Create a batch prediction job over the day's new records
  2. Load the predictions into a results table in Google BigQuery

With Google BigQuery

Apps you can build on Google Vertex AI data

Internal tools your team can build with an AI assistant, each one using Google Vertex AI calls from its API documentation.

Batch job monitor

Lists batch prediction and tuning jobs with their status and lets an engineer cancel a stuck job.

Model endpoint inventory

Lists endpoints and the models on them and lets an engineer undeploy a model that is no longer used.

Prompt cost checker

Counts the tokens in a draft prompt and shows the generated answer so a team can compare prompt versions.

Give your AI agents Google Vertex AI tools

Build an MCP server on Google Vertex AI in Agent Gateway and choose which actions your agents can call. Every call runs with access control and an audit trail. How Agent Gateway works

  • Generate content
  • Create an endpoint
  • Create a batch prediction job
  • Create a tuning job
  • Create cached content
  • Create a dataset
  • Stream generated content
  • Count tokens

What the Google Vertex AI API covers

6 resources and 22 documented actions, each one a call the HTTP Connector can make from a Tray.ai workflow. Connecting it takes a few clicks: here's how.

Publisher models

  • Generate content
  • Stream generated content
  • Count tokens
  • Embed content

Endpoints

  • Create an endpoint
  • List endpoints
  • Deploy a model
  • Get predictions
  • Undeploy a model

Batch prediction jobs

  • Create a batch prediction job
  • List batch prediction jobs
  • Cancel a batch prediction job

Tuning jobs

  • Create a tuning job
  • List tuning jobs
  • Cancel a tuning job

Cached contents

  • Create cached content
  • List cached contents
  • Delete cached content

Datasets

  • Create a dataset
  • Import data
  • Export data
  • List datasets

Checked against Google Vertex AI's own documentation on 2026-10-22

docs.cloud.google.com/vertex-ai/docs/reference/rest

Frequently asked questions

What can a Tray.ai workflow do in Google Vertex AI?

+

The Google Vertex AI API documents 22 actions across 6 resources (Publisher models, Endpoints, Batch prediction jobs, Tuning jobs, Cached contents, Datasets). For example: Generate content; Create an endpoint; Create a batch prediction job. A workflow can make any documented call through the HTTP Connector.

How does Tray.ai authenticate with Google Vertex AI?

+

The Google Vertex AI API uses OAuth 2.0. You save it once as a Tray.ai authentication, and the HTTP Connector adds it to every request.

See Google Vertex AI working against your stack.

We'll walk through a tailored demo with your systems plugged in.

Great — a couple more details

This helps us connect you with the right person. Optional, and you can skip it.