Claude on Google Vertex AI: GCP Integration Guide

Why Claude on Vertex AI?

If your infrastructure runs on Google Cloud, Vertex AI is the most direct path to Claude.

Setup: Step by Step

1. Enable Claude in the Model Garden

1. →Go to the GCP console > Vertex AI > Model Garden
2. →Search for "Claude" in the catalog
3. →Click Enable for the desired models
4. →Accept the Anthropic terms of use

2. Configure a Service Account

# Create a service account
gcloud iam service-accounts create claude-vertex \
    --display-name="Claude Vertex AI"

# Assign the Vertex AI User role
gcloud projects add-iam-policy-binding YOUR_PROJECT_ID \
    --member="serviceAccount:claude-vertex@YOUR_PROJECT_ID.iam.gserviceaccount.com" \
    --role="roles/aiplatform.user"

# Generate a key (for local development)
gcloud iam service-accounts keys create key.json \
    --iam-account=claude-vertex@YOUR_PROJECT_ID.iam.gserviceaccount.com

3. Install the SDKs

pip install anthropic[vertex] google-auth

Code Examples

Basic Call with the Anthropic SDK

import anthropic

client = anthropic.AnthropicVertex(
    project_id="your-gcp-project-id",
    region="us-east5"
)

message = client.messages.create(
    model="claude-sonnet-4@20250514",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Explain the benefits of BigQuery for data analysis."}
    ]
)

print(message.content[0].text)

With System Prompt

message = client.messages.create(
    model="claude-sonnet-4@20250514",
    max_tokens=2048,
    system="You are an expert GCP cloud architect. Recommend Google Cloud solutions.",
    messages=[
        {"role": "user", "content": "I need an architecture for a real-time data pipeline."}
    ]
)

Streaming

with client.messages.stream(
    model="claude-sonnet-4@20250514",
    max_tokens=2048,
    messages=[{"role": "user", "content": "Write a GKE migration guide."}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Authentication on GCP (no JSON key)

# Automatic authentication via Application Default Credentials
# Works on GCE, Cloud Run, GKE, Cloud Functions
client = anthropic.AnthropicVertex(
    project_id="your-project-id",
    region="us-east5"
    # No explicit credentials needed!
)

Pricing: Vertex AI vs Direct API

Vertex pricing advantages:

• →Committed Use Discounts (CUDs): Volume commitment discounts
• →Per-second billing: No minimum
• →GCP credits: Usable for Claude (including free startup credits)
• →Batch predictions: Reduction for batch processing

BigQuery Integration

A powerful use case: using Claude to analyze BigQuery data.

from google.cloud import bigquery
import anthropic

# Retrieve BigQuery data
bq_client = bigquery.Client()
query = """
    SELECT product_name, SUM(revenue) as total_revenue, COUNT(*) as orders
    FROM `project.dataset.sales`
    WHERE date >= DATE_SUB(CURRENT_DATE(), INTERVAL 30 DAY)
    GROUP BY product_name
    ORDER BY total_revenue DESC
    LIMIT 20
"""
results = bq_client.query(query).to_dataframe()

# Analyze with Claude
vertex_client = anthropic.AnthropicVertex(
    project_id="your-project-id",
    region="us-east5"
)

message = vertex_client.messages.create(
    model="claude-sonnet-4@20250514",
    max_tokens=2048,
    messages=[{
        "role": "user",
        "content": f"""Analyze this sales data from the last 30 days:

{results.to_markdown()}

Provide:
1. The top 3 performing products and why
2. Notable trends
3. Recommendations for next month"""
    }]
)

print(message.content[0].text)

Enterprise Architecture on GCP

Typical Architecture

Cloud Run + Claude

# app.py - Cloud Run service with Claude
from flask import Flask, request, jsonify
import anthropic

app = Flask(__name__)
client = anthropic.AnthropicVertex(
    project_id="your-project-id",
    region="us-east5"
)

@app.route("/analyze", methods=["POST"])
def analyze():
    data = request.json
    
    message = client.messages.create(
        model="claude-sonnet-4@20250514",
        max_tokens=2048,
        messages=[{"role": "user", "content": data["question"]}]
    )
    
    return jsonify({
        "answer": message.content[0].text,
        "model": message.model,
        "tokens": {
            "input": message.usage.input_tokens,
            "output": message.usage.output_tokens
        }
    })

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=8080)

FROM python:3.11-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY app.py .
CMD ["python", "app.py"]

# Deploy to Cloud Run
gcloud run deploy claude-service \
    --source . \
    --region us-east5 \
    --service-account claude-vertex@YOUR_PROJECT.iam.gserviceaccount.com

Bedrock vs Vertex AI Comparison

Available Regions

Recommendation: Use europe-west1 for European applications (minimal latency + easier GDPR compliance).

Migrating from the Direct API