Does the Claude exam cover Bedrock or Vertex AI?

The current Claude exam guides do not name Amazon Bedrock or Google Vertex AI as exam objectives. I checked all four guides effective July 2026. Developer – Foundations comes closest: its Claude API Mechanics skill includes invoking Claude through third-party vendors. Architect – Foundations expects deployment-platform judgment, but its guide explicitly excludes specific AWS, GCP and Azure configuration. That is narrower, and more useful, than saying “Bedrock is on the exam.”

Does the Claude exam cover Bedrock? The guide-based answer

There are three different questions hiding inside “cover.” A provider can be named in an objective, a platform choice can be in scope without a vendor name, or a provider-specific course can merely appear in the preparation list. As of 31 August 2026, the first is false, the second is true on relevant technical tracks, and the third is true for Architect – Foundations.

This article cannot promise what wording appears in a live item. Anthropic does not publish the live bank, and its guides say they are subject to change. The defensible answer therefore stops at the current public scope instead of turning one candidate's memory or a local practice set into an exam guarantee. The same rule applies to Claude certification sample questions: use published illustrative items to learn the format, not to predict a live form.

Why the old Bedrock answer was misleading

The previous article said Architect – Foundations had a 19% “Claude API and Model Fundamentals” domain that named Bedrock. That is not the current blueprint. The live Architect – Foundations page now lists five domains led by Agentic Architecture & Orchestration at 27%. Its role summary includes selecting a deployment platform, and its prep catalogue lists separate Amazon Bedrock and Google Cloud courses. The detailed guide, however, names neither provider and puts specific cloud-provider configurations out of scope.

That distinction matters. “Anthropic recommends this course” can support broader preparation; it cannot be converted into “this provider is a named objective,” an item count, or a guarantee that the provider appears on every exam form.

The current scope, exam by exam

This map uses the current guide language, not repository mention counts. A word appearing in a lesson, search index or Cred Farmer practice set does not make it an official objective.

The live pages provide the current high-level blueprints for Associate – Foundations, Developer – Foundations, Architect – Foundations and Architect – Professional. Download the guide from the page you will actually sit and recheck it close to test day.

What cloud-platform knowledge is worth studying?

For Developer – Foundations, “invoking Claude through third-party vendors” sits inside Claude API Mechanics, a 6.8% skill within the 33.1% Applications and Integration domain. The 33.1% is the whole parent domain, not a Bedrock weight. For Architect – Professional, the 19% Integration domain covers authentication and authorisation, observability and protocol selection without naming a cloud vendor.

Study the decisions that transfer across providers: who authenticates the request, where authorisation is enforced, which models and features are available, how regions and data residency constrain a design, who bills the usage, and who operates the inference stack. Anthropic's current feature matrix shows why the route is not cosmetic: feature availability can differ between the direct Claude API, Amazon Bedrock and Google Cloud.

Do you need an AWS or Google Cloud account?

Not to satisfy a published exam prerequisite. No current guide lists a cloud account or performance lab as a prerequisite, and the exam format is multiple choice and multiple response. Three guides explicitly say there is no mandatory course; Architect – Foundations publishes recommended experience but no cloud account requirement. Hands-on work can help, but you do not bring that account into the exam.

An account can still make optional practice concrete. If a measured gap is Bedrock request structure or Vertex endpoint selection, use the current Claude in Amazon Bedrock or Claude on Vertex AI documentation, then run only the exercise you need. For conceptual platform selection, a comparison table and a written design rationale may be enough; do not create billable infrastructure merely to memorise a console path that the guide excludes.

Key takeaways

• No current guide names Bedrock or Vertex AI. That is a statement about the four published July 2026 guides, not a claim about unpublished live-item wording.
• Developer – Foundations has the clearest generic coverage. Its API Mechanics skill includes invoking Claude through third-party vendors.
• Architect – Foundations tests platform judgment, not provider setup. Its public page mentions deployment-platform selection; the guide excludes specific AWS/GCP/Azure configuration.
• A prep-course listing is not an exam objective. Treat courses as broader preparation, not proof of a named provider item.
• Learn portable decisions first. Authentication, feature availability, regions, data residency, operating responsibility and billing survive provider changes better than console steps.