SQL++ 및 MCP(Model Context Protocol)를 사용하여 AI 에이전트에 실시간 액세스 권한 부여
자율형 AI 에이전트가 더욱 강력해지고 널리 보급됨에 따라, 개발자들은 이러한 에이전트를 운영 및 분석 데이터에 안전하고 안정적으로 연결할 수 있는 방법이 필요합니다. 고객 서비스 에이전트, 지능형 대시보드, 에이전트 워크플로 중 무엇을 구축하든 데이터 Access는 지능의 핵심입니다.
그렇기 때문에 기쁜 마음으로 발표합니다 카우치베이스는 이제 공식적으로 다음에서 지원됩니다 Google MCP 데이터베이스용 툴박스.

이 통합은 Couchbase의 고성능 및 유연한 NoSQL 데이터베이스와 도구 기반 에이전트 오케스트레이션을 위한 MCP 표준 사이의 간격을 좁혀 개발자가 에이전트형 AI 애플리케이션의 개발을 가속화할 수 있도록 지원합니다. MCP 툴박스 내에서 Couchbase를 활용함으로써 팀은 안전하고 확장 가능하며 컨텍스트를 인식하는 지능형 데이터 기반 워크플로우를 구현할 수 있습니다. 개발자는 더 이상 커스텀 커넥터를 구축하거나 복잡한 액세스 로직을 관리하는 데 시간을 낭비할 필요가 없습니다. 이제 최소한의 마찰로 운영 데이터를 조회하고 이에 따라 조치를 취하는 더 스마트한 에이전트를 구축하는 데 집중할 수 있습니다. 이번 통합은 Couchbase의 성능, 유연성 및 실시간 기능을 새로 부상하는 모델 컨텍스트 프로토콜(MCP) 호환 AI 도구 생태계로 가져다줍니다.
개발자에게 Couchbase + MCP Toolbox가 중요한 이유
데이터베이스용 Google MCP 툴박스는 MCP 서버, 데이터베이스를 직접 노출하거나 각 유스케이스마다 커스텀 통합 코드를 작성할 필요 없이, AI 에이전트가 선언적 도구 정의를 통해 정형 데이터와 상호작용할 수 있도록 합니다.
Couchbase 지원을 추가함으로써 개발자는 다음과 같은 이점을 얻을 수 있습니다:
- 서식 문서 모델링 하위 문서 액세스 및 유연한 스키마를 통해.
- 내장된 확장성과 동기화 엣지, 모바일 및 분산 애플리케이션용.
- MCP 호환성, AI 에이전트가 안전하고 표준화된 도구 인터페이스를 통해 Couchbase 데이터를 추론하고 쿼리할 수 있도록 합니다.
이제 에이전트는 하드코딩된 로직이나 커스텀 API 생성, 데이터베이스 자격 증명 노출 없이, 이면에서 SQL++ 도구 호출로 변환되는 자연어 쿼리를 실행할 수 있습니다.
작동 방식
그 Google MCP 데이터베이스용 툴박스 랭체인(LangChain), 랭그래프(LangGraph) 또는 앤스로픽의 클로드(Anthropic's Claude) 같은 AI 오케스트레이션 레이어를 데이터베이스 기반 도구에 연결해 주는 오픈소스 MCP 서버입니다.
Couchbase 접근을 구성하려면 YAML 설정 파일에 데이터 소스와 도구를 정의합니다.
샘플 tools.yaml
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
출처: cb-주문: 친절한카우치베이스 연결 문자열카우치베이스://localhost:8091 양동이주문_버킷 범위주문 사용자 이름툴박스 사용자 비밀번호: $CB_PASSWORD 도구: 고객 주문 조회: 설명이메일로 특정 고객의 최근 주문 조회하기. 원천: cb-orders 질의: | 선택 주문_아이디, 전체, 상태, 주문_날짜 ~에서 주문_양동이 어디 고객.이메일 = $이메일 주문 에 의해 주문_날짜 설명 한계 10; 매개변수: – 이름이메일 유형: string 설명: customer email |
Once deployed, your AI agents can invoke 고객 주문 조회 by passing the 이메일 parameter, and the MCP Toolbox will translate the request into a secure SQL++ query against Couchbase.
Agentic use cases for Couchbase + MCP
Conversational BI dashboards
Build agents that generate dynamic dashboards and metrics by querying Couchbase for business insights. Couchbase’s SQL++ support allows complex aggregations and JOINs, even across flexible schemas.
- 예시: “What were the top 10 product categories last quarter by revenue?”
E-Commerce Agent Assistants
Serve product recommendations, inventory checks, and order histories from JSON documents in Couchbase. Support fast lookups and full-text search to guide users in real-time.
- Example: “Show me available trail running shoes under $120 in size 11.”
IoT and Edge Analytics
Couchbase Mobile and Sync Gateway bring real-time data from edge devices into Couchbase. Agents can analyze telemetry, spot anomalies, or initiate remediation workflows using SQL++ queries.
- 예시: “List sensors with temperature fluctuations over 10 degrees in the past hour.”
Secure Customer Copilots
Build personalized agents that help users understand their own data—finances, activity, healthcare, etc.—by securely scoping SQL++ queries to their identity context.
- 예시: “Summarize my recent transactions and categorize them.”
LangChain + Couchbase via MCP: integration example
Here’s how you can wire Couchbase tools into a LangChain agent via MCP Toolbox:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 |
에서 langgraph.prebuilt 가져오기 create_react_agent 에서 langchain_openai 가져오기 ChatOpenAI MCP_ENDPOINT = “https://localhost:8080” tool_box = ToolboxClient(MCP_ENDPOINT) toolbox_tools = toolbox_client.load_toolset(“toolset_name”) 대규모 언어 모델 = ChatOpenAI( 모델=“gpt-4o”, temperature=0.1, API 키=settings.openai_api_열쇠 ) 에이전트 = create_react_agent( 도구=toolbox_tools, 모델=대규모 언어 모델, 디버그=True ) 에이전트.invoke( {“messages”: [{“role”: “user”, “content”: “Get the last few orders for alice@example.com.”}]} ) |
Built for production
The MCP Toolbox for Databases offers production-grade features out of the box:
- Config-driven deployment with zero downtime
- Connection pooling and retry logic
- OAuth2 / OIDC auth for secure access
- OpenTelemetry support for metrics and traces
- Support for multiple databases from a single config
You can deploy the toolbox as a microservice alongside your orchestration layer (e.g. in GKE, Cloud Run, or Kubernetes Anywhere).
Get started
- Google MCP Toolbox for Databases (GitHub)
- 카우치베이스 개발자 포털
- Deploy locally or in the cloud with Docker, Cloud Run, or Kubernetes
In summary
With this new integration, Couchbase joins the growing list of production-ready backends in the Google MCP ecosystem. Developers now have a fast, scalable, and secure way to enable AI agents to access NoSQL document data using standardized protocols.
This lowers the barrier to building intelligent applications that react in real-time to user inputs, contextual data, and live system states, all powered by Couchbase and governed by the open MCP standard.

댓글 남기기
댓글을 달기 위해서는 로그인해야합니다.