Exam glossary
Google Cloud Agentic Architect glossary
The terms you meet in the Google Cloud Agentic Architect exam, with a short definition. The exam is in English: the German equivalent helps if you study in German. Portuguese, Spanish and French will follow.
27 terms · Definitions checked against official documentation on November 13, 2026 · Auf Deutsch
A
- Agent as a tool#German: Agent als Werkzeug
- Calling another agent like a function and getting its result back, while the calling agent keeps control. With sub-agent delegation, control passes to the sub-agent.
- Exam domain: Developing custom agents
- Agent Development Kit(ADK)#German: Agent Development Kit (ADK)
- Google's open-source framework for building agents and multi-agent systems in code.
- Exam domain: Developing custom agents
- Agent evaluation#German: Agenten-Evaluierung
- Measuring an agent's quality, including whether it chose the right tools and steps (trajectory), not only whether the final answer is correct.
- Exam domain: Evaluating and deploying agentic workflows
- Agent Gateway#German: Agent Gateway
- A control point for traffic between agents, tools and models, used for policy enforcement and monitoring.
- Exam domain: Securing and governing agentic workflows
- Agent Identity#German: Agent Identity
- A dedicated identity for an agent so that its access can be controlled and audited with IAM.
- Exam domain: Securing and governing agentic workflows
- Agent Registry#German: Agent Registry
- A central catalogue of agents, tools and MCP servers in an organisation, used for discovery and governance.
- Exam domain: Securing and governing agentic workflows
- Agent Runtime#German: Agent Runtime
- The managed service for deploying and running agents in production, formerly Vertex AI Agent Engine.
- Exam domain: Evaluating and deploying agentic workflows
- Agent Search#German: Agent Search
- Google's managed enterprise search for grounding agents, formerly Vertex AI Search.
- Exam domain: Developing custom agents
- Agent Studio#German: Agent Studio
- The console environment for prototyping prompts, models and agents, formerly Vertex AI Studio.
- Exam domain: Building agents using low-code tools
- Agent2Agent protocol(A2A)#German: Agent2Agent-Protokoll (A2A)
- An open protocol that lets agents built with different frameworks discover each other and work together.
- Exam domain: Developing custom agents
- Agentic workflow#German: Agentischer Workflow
- A process in which one or more agents carry out steps, call tools and pass results to each other, with or without fixed control flow.
- Exam domain: Developing custom agents
- AI agent#German: KI-Agent
- A system that uses a model to reason about a goal, plan steps and call tools to carry them out.
- Exam domain: Developing custom agents
C
- Callbacks#German: Callbacks (Rückruffunktionen)
- Code that ADK runs before or after a model or tool call, used for example for guardrails, logging or changing requests.
- Exam domain: Securing and governing agentic workflows
- Coding agent#German: Coding-Agent
- An agent that reads, writes and runs code in a repository. The exam guide names Antigravity and Claude Code on Google Cloud.
- Exam domain: Using coding agents for application development
- CX Agent Studio#German: CX Agent Studio
- Google's low-code tool for building customer-experience agents, such as service chat and voice agents.
- Exam domain: Building agents using low-code tools
G
- Gemini Enterprise Agent Platform#German: Gemini Enterprise Agent Platform
- Google Cloud's platform for building and running agents and models; the new name for what was Vertex AI.
- Exam domain: Evaluating and deploying agentic workflows
- Grounding#German: Grounding (Fundierung)
- Connecting model output to trusted sources, such as enterprise data or search results, to reduce made-up answers.
- Exam domain: Developing custom agents
M
- Memory Bank#German: Memory Bank
- Long-term memory across sessions, so an agent can recall facts about a user in later conversations.
- Exam domain: Evaluating and deploying agentic workflows
- Model Armor#German: Model Armor
- A Google Cloud service that screens prompts and responses for risks such as prompt injection, harmful content and sensitive data.
- Exam domain: Securing and governing agentic workflows
- Model Context Protocol(MCP)#German: Model Context Protocol (MCP)
- An open protocol for connecting agents to tools and data through MCP servers.
- Exam domain: Using coding agents for application development
- Multi-agent system#German: Multi-Agenten-System
- Several specialised agents that work together, for example a coordinator that delegates tasks to sub-agents.
- Exam domain: Developing custom agents
O
- OAuth 2.0#German: OAuth 2.0
- The authorisation standard used when an agent calls a tool on behalf of a user, so the agent acts with the user's delegated permissions.
- Exam domain: Securing and governing agentic workflows
R
- RAG Engine#German: RAG Engine
- A managed Google Cloud service for building RAG pipelines: ingesting, chunking, embedding and retrieving documents.
- Exam domain: Developing custom agents
- Retrieval-augmented generation(RAG)#German: Retrieval-Augmented Generation (RAG)
- Retrieving relevant enterprise content and adding it to the model's context so answers are grounded in that data.
- Exam domain: Developing custom agents
S
- Sequential, parallel and loop agents#German: Sequenzielle, parallele und Schleifen-Agenten
- ADK workflow agents with fixed control flow: run sub-agents one after another, at the same time, or repeatedly until a condition is met.
- Exam domain: Developing custom agents
- Sessions#German: Sessions (Sitzungen)
- Storage for the conversation history and state of one interaction with an agent (Agent Platform Sessions).
- Exam domain: Evaluating and deploying agentic workflows
W
- Workflow Builder#German: Workflow Builder
- The low-code tool in Gemini Enterprise for building agents without code; the exam guide still calls it Agent Designer.
- Exam domain: Building agents using low-code tools