Vertex AI Agent Engine(이전 명칭: Vertex AI 기반 LangChain 또는 Vertex AI Reasoning Engine)은 개발자가 프로덕션에서 AI 에이전트를 배포, 관리, 확장할 수 있는 완전 관리형 Google Cloud 서비스입니다. Agent Engine은 인프라를 처리하여 프로덕션에서 상담사를 확장하므로 개발자는 지능적이고 강력한 애플리케이션을 만드는 데 집중할 수 있습니다. Vertex AI Agent Engine은 다음을 제공합니다.
완전 관리형: VPC-SC 규정 준수 및 포괄적인 엔드 투 엔드 관리 기능을 포함한 강력한 보안 기능을 제공하는 관리형 런타임으로 에이전트를 배포하고 확장합니다. 성능 모니터링 및 트레이스에 Google Cloud Trace(OpenTelemetry 지원)를 사용하는 멀티 에이전트 애플리케이션에 대한 CRUD 액세스 권한을 얻습니다. 자세한 내용은 에이전트 배포를 참조하세요.
간소화된 개발: Vertex AI Agent Engine은 애플리케이션 서버 개발, 인증 및 IAM 구성과 같은 하위 수준 태스크를 추상화하므로 동작, 도구, 모델 파라미터와 같은 에이전트의 고유한 기능에 집중할 수 있습니다. 또한 에이전트는 Vertex AI에서 함수 호출과 같은 모든 모델과 도구를 사용할 수 있습니다.
프레임워크 독립형: LangGraph, Langchain, AG2, CrewAI를 포함한 다양한 Python 프레임워크를 사용하여 빌드한 에이전트를 유연하게 배포할 수 있습니다.
기존 에이전트가 이미 있는 경우 SDK의 커스텀 템플릿을 사용하여 Vertex AI Agent Engine에서 실행되도록 조정할 수 있습니다. 또는 Google에서 제공하는 프레임워크별 템플릿 중 하나를 사용하여 에이전트를 처음부터 개발할 수 있습니다.
Vertex AI Agent Engine에서 만들기 및 배포
참고: Vertex AI Agent Engine으로 간소화된 IDE 기반 개발 및 배포 환경을 사용하려면 agent-starter-pack을 고려하세요. 즉시 사용할 수 있는 템플릿, 실험을 위한 기본 제공 UI를 제공하고 배포, 운영, 평가, 맞춤설정, 관찰 가능성을 간소화합니다.
Vertex AI Agent Engine에서 에이전트를 빌드하는 워크플로는 다음과 같습니다.
Vertex AI Agent Engine은 데이터 보안이 강화되고 데이터 무단 반출 위험이 완화되도록 VPC 서비스 제어를 지원합니다. VPC 서비스 제어가 구성되면 배포된 에이전트는 BigQuery API, Cloud SQL Admin API, Vertex AI API와 같은 Google API 및 서비스에 대한 보안 액세스 권한을 유지하므로 정의된 경계 내에서 원활한 작업을 보장합니다. 중요한 점은 VPC 서비스 제어는 모든 공개 인터넷 액세스를 효과적으로 차단하여 승인된 네트워크 경계로 데이터 이동을 제한하고 엔터프라이즈 보안 상황을 크게 향상시킨다는 점입니다.
[[["이해하기 쉬움","easyToUnderstand","thumb-up"],["문제가 해결됨","solvedMyProblem","thumb-up"],["기타","otherUp","thumb-up"]],[["이해하기 어려움","hardToUnderstand","thumb-down"],["잘못된 정보 또는 샘플 코드","incorrectInformationOrSampleCode","thumb-down"],["필요한 정보/샘플이 없음","missingTheInformationSamplesINeed","thumb-down"],["번역 문제","translationIssue","thumb-down"],["기타","otherDown","thumb-down"]],["최종 업데이트: 2025-09-03(UTC)"],[],[],null,["# Vertex AI Agent Engine overview\n\n| [VPC-SC security controls](/vertex-ai/docs/general/vpc-service-controls)\n| and [Customer-managed encryption keys (CMEK)](/vertex-ai/docs/general/cmek) are\n| supported by Vertex AI Agent Engine. Data residency (DRZ) and Access Transparency (AXT)\nsecurity controls aren't supported. \n| To see an example of getting started with Vertex AI Agent Engine,\n| run the \"Building and Deploying an Agent with \" notebook in one of the following\n| environments:\n|\n| [Open in Colab](https://colab.research.google.com/github/GoogleCloudPlatform/generative-ai/blob/main/gemini/agent-engine/intro_agent_engine.ipynb)\n|\n|\n| \\|\n|\n| [Open in Colab Enterprise](https://console.cloud.google.com/vertex-ai/colab/import/https%3A%2F%2Fraw.githubusercontent.com%2FGoogleCloudPlatform%2Fgenerative-ai%2Fmain%2Fgemini%2Fagent-engine%2Fintro_agent_engine.ipynb)\n|\n|\n| \\|\n|\n| [Open\n| in Vertex AI Workbench](https://console.cloud.google.com/vertex-ai/workbench/deploy-notebook?download_url=https%3A%2F%2Fraw.githubusercontent.com%2FGoogleCloudPlatform%2Fgenerative-ai%2Fmain%2Fgemini%2Fagent-engine%2Fintro_agent_engine.ipynb)\n|\n|\n| \\|\n|\n| [View on GitHub](https://github.com/GoogleCloudPlatform/generative-ai/blob/main/gemini/agent-engine/intro_agent_engine.ipynb)\n\nVertex AI Agent Engine, a part of the Vertex AI Platform, is a\nset of services that enables developers to deploy, manage, and scale AI agents\nin production. Agent Engine handles the infrastructure to scale agents in\nproduction so you can focus on creating applications.\nVertex AI Agent Engine offers the following services that you can\nuse individually or in combination:\n\n- **Runtime**:\n\n - [Deploy](/vertex-ai/generative-ai/docs/agent-engine/deploy) and scale agents with a managed runtime and end-to-end management capabilities.\n - Customize the agent's container image with build-time installation scripts for system dependencies.\n - Use security features including VPC-SC compliance and configuration of authentication and IAM.\n - Access models and tools such as [function calling](/vertex-ai/generative-ai/docs/multimodal/function-calling).\n - Deploy agents built using [different Python frameworks](#supported-frameworks):\n - Understand agent behavior with [Google\n Cloud Trace](/vertex-ai/generative-ai/docs/agent-engine/manage/tracing) (supporting [OpenTelemetry](https://opentelemetry.io/)), [Cloud Monitoring](/vertex-ai/generative-ai/docs/agent-engine/manage/monitoring), and [Cloud Logging](/vertex-ai/generative-ai/docs/agent-engine/manage/logging).\n- **Quality and evaluation** (Preview): Evaluate agent quality with the integrated\n [Gen AI Evaluation service](/vertex-ai/generative-ai/docs/agent-engine/evaluate) and optimize agents with Gemini model training runs.\n\n- [**Example Store**](/vertex-ai/generative-ai/docs/example-store/overview) (Preview): Store and dynamically retrieve few-shot examples to improve agent performance.\n\n- [**Sessions**](/vertex-ai/generative-ai/docs/agent-engine/sessions/overview) (Preview): Agent Engine Sessions lets you store individual interactions between users and agents, providing definitive sources for conversation context.\n\n- [**Memory Bank**](/vertex-ai/generative-ai/docs/agent-engine/memory-bank/overview) (Preview): Agent Engine Memory Bank lets you store and retrieve information from sessions to personalize agent interactions.\n\nVertex AI Agent Engine is part of [Vertex AI Agent Builder](/vertex-ai/generative-ai/docs/agent-builder/overview), a suite of features for discovering, building, and deploying AI agents.\n| **Note:** Because the name of Vertex AI Agent Engine changed over time, the name of the resource in the API reference is [`ReasoningEngine`](/vertex-ai/generative-ai/docs/reference/rest/v1/projects.locations.reasoningEngines) to maintain backwards compatibility.\n\nCreate and deploy on Vertex AI Agent Engine\n-------------------------------------------\n\n**Note:** For a streamlined, *IDE-based* development and deployment experience with Vertex AI Agent Engine, consider the [agent-starter-pack](https://github.com/GoogleCloudPlatform/agent-starter-pack). It provides ready-to-use templates, a built-in UI for experimentation, and simplifies deployment, operations, evaluation, customization, and observability.\n\nThe workflow for building an agent on Vertex AI Agent Engine is:\n\nThe steps are illustrated by the following diagram:\n\n\u003cbr /\u003e\n\nSupported frameworks\n--------------------\n\nThe following table describes the level of support Vertex AI Agent Engine provides for various agent frameworks:\n\nDeploy in production with Agent Starter Pack\n--------------------------------------------\n\nThe [Agent Starter Pack](https://github.com/GoogleCloudPlatform/agent-starter-pack) is a collection of production-ready generative AI agent templates built for Vertex AI Agent Engine. The Agent Starter Pack provides the following:\n\n- **Pre-built agent templates:** ReAct, RAG, multi-agent, and other templates.\n- **Interactive playground**: Test and interact with your agent.\n- **Automated infrastructure** : Uses [Terraform](https://cloud.google.com/docs/terraform) for streamlined resource management.\n- **CI/CD pipelines**: Automated deployment workflows leveraging Cloud Build.\n- **Observability**: Built-in support for Cloud Trace and Cloud Logging.\n\nTo get started, see the [Quickstart](https://github.com/GoogleCloudPlatform/agent-starter-pack?tab=readme-ov-file#-get-started-in-1-minute).\n\nUse cases\n---------\n\nTo learn about Vertex AI Agent Engine with end-to-end examples, see the following\nresources:\n\nEnterprise security\n-------------------\n\nVertex AI Agent Engine supports several features to help you meet\nenterprise security requirements, adhere to your organization's security\npolicies, and follow security best practices. The following features are\nsupported:\n\n- **Customer-managed encryption keys (CMEK)** : Vertex AI Agent Engine\n supports [CMEK](/kms/docs/cmek) to protect your data with your own encryption keys,\n which gives you ownership and full control of the keys that protect your data at\n rest in Google Cloud. For more information, see\n [Agent Engine CMEK](/vertex-ai/generative-ai/docs/agent-engine/manage/access#cmek).\n\n- **VPC Service Controls** : Vertex AI Agent Engine supports\n [VPC Service Controls](/vertex-ai/docs/general/vpc-service-controls) to strengthen data\n security and mitigate the risks of data exfiltration. When VPC Service Controls\n is configured, the deployed agent retains secure access to Google APIs and\n services, such as BigQuery API, Cloud SQL Admin API, and Vertex AI API,\n verifying seamless operation within your defined perimeter. Critically,\n VPC Service Controls effectively blocks all public internet access, confining\n data movement to your authorized network boundaries and significantly\n enhancing your enterprise security posture.\n\n | **Note:** If you use Vertex AI Agent Engine in a VPC-SC environment, you must create an ingress rule in your perimeter to allow ingress from the Reasoning Engine Service Agent (`service-`\u003cvar translate=\"no\"\u003ePROJECT_NUMBER\u003c/var\u003e`@gcp-sa-aiplatform-re.iam.gserviceaccount.com`) to `storage.googleapis.com` service and `artifactregistry.googleapis.com` service.\n- **Private Service Connect interface** : [PSC-I](/vpc/docs/about-private-service-connect-interfaces)\n lets your agents interact with privately hosted services in user's VPC. For\n more information, see [Configure VPC Service Controls interface](/vertex-ai/generative-ai/docs/agent-engine/deploy#psc-i).\n\n- **HIPAA** : As a part of Vertex AI Platform,\n Vertex AI Agent Engine supports [HIPAA](/security/compliance/hipaa)\n workloads.\n\nSupported regions\n-----------------\n\nVertex AI Agent Engine is supported in the following regions:\n\nFor [Agent Engine Memory Bank](/vertex-ai/generative-ai/docs/agent-engine/memory-bank/overview) (Preview), the following regions are supported:\n\nQuota\n-----\n\nThe following limits apply to [Vertex AI Agent Engine](/vertex-ai/generative-ai/docs/agent-engine/overview) for a given project in each region: \n\nPricing\n-------\n\nThe pricing for the Agent Engine Runtime is based on\nthe compute (vCPU hours) and memory (GiB hours) your agent uses to process\nrequests. There is no charge for the time a deployed agent is idle.\n\nFor more information, see [pricing](/vertex-ai/pricing#agent_engine).\n\nWhat's next\n-----------\n\n- [Set up the environment](/vertex-ai/generative-ai/docs/agent-engine/set-up).\n- [Get support](/vertex-ai/generative-ai/docs/agent-engine/support)."]]