Claude, governed like the rest of your Google Cloud
Models change quickly. The one that leads on code this quarter may not lead on reasoning next quarter, so tying your architecture to a single model provider is a risk.
We help Australian organisations run Anthropic’s Claude inside their own Google Cloud Vertex AI tenant. You get Claude’s reasoning, coding and tool use under the IAM, encryption and billing controls you already run.
Why run Claude on Vertex AI
Risk committees often block direct API subscriptions to a new vendor. Running Claude through Vertex AI removes most of the objections.
Your data stays under Google Cloud terms. Prompts and outputs are handled under your Google Cloud agreement, and your data is not used to train Anthropic’s models.
One set of controls and one bill. No separate vendor contract and no API keys on developer laptops. Claude usage appears on your Google Cloud bill and is governed by your existing IAM policies, customer-managed encryption keys (CMEK) and VPC Service Controls.
Connected to your systems through MCP. Using the Model Context Protocol (MCP), we connect Claude agents to BigQuery, Cloud SQL, Jira, GitHub and internal services, so Claude can take action rather than just produce text.
How the architecture fits together
- Orchestration: Agent Development Kit (ADK) or LangGraph running on Cloud Run.
- Reasoning: Claude on Vertex AI, inside your Google Cloud tenant.
- Tools: MCP servers giving controlled access to BigQuery, Cloud SQL, Jira and internal APIs.
What we build with it
Multi-agent systems. Claude handles planning, code changes and synthesis, while Gemini models handle high volume multimodal and document processing.
Code modernisation pipelines. Automated pipelines that analyse legacy codebases, generate regression tests and refactor sections into TypeScript or Go.
Contact centre assist. Claude summarises conversations, checks responses against policy documents and drafts replies grounded in your knowledge base.
How we deliver
We write the Terraform, configure the Vertex AI endpoints and deploy the agent services into your environment.
- Security and governance: VPC Service Controls, DLP masking of personal information, and IAM roles so developers can use Claude safely.
- MCP servers: connecting Claude to your databases and APIs.
- Evaluation: golden datasets and CI/CD gates that measure accuracy, latency and cost before release.
- Handover: we pair with your engineers so you own the code, prompts and orchestration.
Book a call or see our agentic engineering services.
Talk to an architect who has done this before.
Bring your current setup and the outcome you need. You will get a view on the approach, the risks and roughly what it costs.
Straight to a senior GCP architect. No SDR, no slide deck.
Not ready to talk? See how we migrated Hapana off AWS →
Or call +61 2 8359 9507 · Hello@aviato.consulting