> For the complete documentation index, see [llms.txt](https://docs.flxbl.io/flxbl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.flxbl.io/flxbl/sfp-server/configuring-ai-providers/claude-on-vertex-ai.md).

# Claude on Vertex AI

codev's AI features — architecture analysis, validation error analysis, and AI-assisted reviews — run through an AI provider you connect once, in the codev UI. You don't edit config files or set environment variables by hand; you add the provider under **Settings → Integrations** and codev wires it into the analysis engine for you.

This page covers the AI providers available today and where **Claude on Google Cloud Vertex AI** fits in.

## Connect an AI provider in codev

1. In codev, open **Settings → Integrations**.
2. Find the **AI Providers** section.
3. Pick a provider card and click **Connect**.
4. Follow the in-dialog instructions to paste your credential, then **Save**.

<figure><img src="/files/41x3naRkwF7RqIf00AZC" alt="The AI Providers section of Settings -> Integrations, showing provider cards including Anthropic, OpenAI, Google AI, GitHub Copilot, Amazon Bedrock, and Claude on Google Vertex"><figcaption><p>The <strong>AI Providers</strong> section under <strong>Settings → Integrations</strong>. Each card connects one provider.</p></figcaption></figure>

That's it — the next AI command (for example an architecture analysis from the **Analysis** page) uses the connected provider. To confirm it works, run an analysis and check that a real result comes back.

### Available today

| Provider                    | What you provide                                                | Where to get it                                                   |
| --------------------------- | --------------------------------------------------------------- | ----------------------------------------------------------------- |
| **Anthropic (Claude)**      | API key                                                         | [console.anthropic.com](https://console.anthropic.com) → API Keys |
| **OpenAI**                  | API key                                                         | [platform.openai.com](https://platform.openai.com) → API keys     |
| **Google AI (Gemini)**      | API key                                                         | [aistudio.google.com](https://aistudio.google.com) → Get API key  |
| **GitHub Copilot**          | Token (active Copilot subscription)                             | GitHub → Copilot                                                  |
| **Amazon Bedrock**          | AWS credentials                                                 | AWS console                                                       |
| **Claude on Google Vertex** | Google Cloud project ID, region, and a service-account JSON key | Google Cloud → Vertex AI (see below)                              |

{% hint style="info" %}
If you already have a direct **Anthropic API key**, the **Anthropic (Claude)** card is the fastest way to use Claude — connect it and you're done. The Vertex path below is for organizations that have licensed Claude through Google Cloud and want model traffic and billing to stay inside their Google Cloud project.
{% endhint %}

## Claude on Vertex AI

Many organizations license **Claude through Google Cloud Vertex AI** (Model Garden) rather than the direct Anthropic API, so that model traffic, billing, and data handling stay governed by their existing Google Cloud agreement.

Connecting the **Claude on Google Vertex** card has two parts: first you get Claude working in your **Google Cloud project** (one-time Google-side setup), then you **connect it in codev**.

### Step 1 — Get Claude working in your Google Cloud project

Claude on Vertex is gated by **two independent things**, and you need **both**. Enabling the model does *not* grant the quota to call it — these are separate systems.

**1a. Enable the model in Model Garden (accept Anthropic's terms).**

* In the Google Cloud Console, make sure the project has **billing enabled** and the **Vertex AI API** turned on.
* Go to **Vertex AI → Model Garden**, open a Claude model (e.g. Claude Opus), and click **Enable**.
* This opens a short **access request form** (your business name, website, contact, region, industry, intended users, and use case). Anthropic reviews access requests, so the information is shared with Anthropic to approve your project.
* After the form, you accept the **Google Cloud Marketplace + Anthropic Terms of Service** on a pricing/agreement page (Claude on Vertex is pay-as-you-go — you're billed per token only when you use it, with no upfront charge). When it finishes you'll see **"Successfully purchased"**.

<figure><img src="/files/WsluIV3fyFQL1oPtRce5" alt="The Claude Opus model page in Vertex AI Model Garden, showing the Enable button and the Model ID"><figcaption><p>Open a Claude model in Vertex AI Model Garden and click <strong>Enable</strong> to start the access request + terms acceptance. codev selects the model for you — you never set a model ID in the connect form.</p></figcaption></figure>

**1b. Make sure you have quota for the model.**

{% hint style="danger" %}
**Enabling the model is not enough on its own.** A new Google Cloud project starts with a **default quota of `0`** for partner (Anthropic) models, so calls fail with **HTTP 429 "Quota exceeded ... base model: `anthropic-claude-<model>`"** even right after you've accepted the terms. Projects that already use Claude on Vertex usually have quota; a brand-new project does not.

If your project shows `0` quota, request an increase for the `online_prediction_requests_per_base_model` metric (filter by your Claude `base_model`) via **IAM & Admin → Quotas**. The self-service form can be **auto-declined** on a brand-new billing account that has no real payment history. If that happens, the reliable path is a **Google Cloud Support case** (under **Vertex AI → Generative AI → Model access / Quota**) including your project ID, billing account ID, and your expected requests-per-minute and tokens-per-minute — Google's capacity-planning team reviews it (typically a couple of business days). An organization that already licenses Claude through Vertex normally has this in place already.
{% endhint %}

**Other Google-side prerequisites:**

* A **region** of `us` or `eu` (Claude on Vertex is served from these multi-region endpoints).
* **Credentials**: a service account with the **Vertex AI User** role (`roles/aiplatform.user`), or a workload identity if codev runs on Google Cloud compute. Claude on Vertex uses Google IAM auth, not an API key.

{% hint style="info" %}
Anthropic's own [Claude on Vertex AI guide](https://code.claude.com/docs/en/google-vertex-ai) is the authoritative reference for which Claude models and regions are available and how Vertex auth and quota work.
{% endhint %}

### Step 2 — Connect it in codev

Once Claude is enabled and has quota in your project, connect it like any other provider: open **Settings → Integrations → AI Providers**, pick the **Claude on Google Vertex** card, click **Connect**, then paste your **Google Cloud project ID**, **region** (`us` or `eu`), and a **service-account JSON key**, and **Save**.

<figure><img src="/files/cQ65RoTMIoUuDyd08rvW" alt="The Claude on Google Vertex connect form showing fields for Google Cloud Project ID, Region, and Service Account Key (JSON)"><figcaption><p>The <strong>Claude on Google Vertex</strong> connect form: project ID, region (<code>us</code> or <code>eu</code>), and a service-account JSON key. codev picks the latest available Claude (Opus, then Sonnet) automatically — there's no model field to set.</p></figcaption></figure>

## Verify

Once an AI provider is connected, run any AI command that uses it — an **architecture analysis** from the Analysis page is a good check — and confirm a real model response comes back. For Vertex specifically, a successful round-trip confirms the project, region, credentials, model ID, **and** quota are all correct.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.flxbl.io/flxbl/sfp-server/configuring-ai-providers/claude-on-vertex-ai.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.