Run:AI MCP Server
Provides tools to control NVIDIA Run:AI infrastructure, including searching and calling Run:AI REST APIs, resolving entity names to IDs, and executing CLI commands for workload management, diagnostics, port-forwarding, and logs.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Run:AI MCP Server show me the status of my training workloads"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Run:AI MCP Server
NVIDIA Run:AI 인프라를 LLM 에이전트가 제어하는 MCP(Model Context Protocol) 서버.
426개 Run:AI API를 4개 메타 도구 + BM25 검색 엔진으로 동적 검색·실행합니다.
도구
도구 | 설명 |
| OpenAPI 스펙 BM25 검색 — 엔드포인트, 메서드, 스키마 반환 |
| REST API 호출 (인증 자동, DELETE 시 confirm 필요) |
| 엔티티 이름 → ID 변환 (cluster, project, workload 등) |
| CLI 전용 작업 (port-forward, exec, logs, attach, diagnostics) |
Related MCP server: nvidia-nim-mcp
워크플로우
search_runai_api_spec("create training workload")
→ resolve_runai_names([{ type: "project", name: "my-project" }])
→ call_runai_api({ method: "POST", path: "/api/v1/workloads/trainings", ... })설치
npm install
npm run build인증
Run:AI UI에서 Application (서비스 계정) 생성 → clientID + clientSecret 발급.
서버가 자동으로 POST /api/v1/token (client_credentials grant)으로 JWT를 발급·갱신합니다.
cp .env.example .env
# RUNAI_CLIENT_ID, RUNAI_CLIENT_SECRET 설정정적 토큰(RUNAI_API_TOKEN)도 지원하나, JWT는 만료되므로 서비스 계정 방식을 권장합니다.
Claude Code MCP 설정
~/.claude/mcp.json:
{
"mcpServers": {
"runai": {
"command": "node",
"args": ["/path/to/runai-mcp-server/dist/index.js"],
"env": {
"RUNAI_API_BASE_URL": "https://your-instance.run.ai",
"RUNAI_CLIENT_ID": "your-client-id",
"RUNAI_CLIENT_SECRET": "your-client-secret"
}
}
}
}개발
npm run dev # ts-node 개발 실행
npm test # 전체 테스트
npm run build # 프로덕션 빌드아키텍처
Registry Builder: OpenAPI 3.0.3 스펙을 파싱하여 413개 OperationDescriptor 생성, $ref 자동 해결
BM25 Search: IDF 기반 키워드 검색 + 동의어 확장 (job→workload, remove→delete 등)
Adaptive Schema: 검색 결과 1-2개면 requestBody 스키마 포함, 3개+면 summary만 (토큰 절약)
Safety: DELETE는 confirm_destructive 필수, CLI는 화이트리스트 + 쉘 인젝션 방어
라이선스
MIT
Available Tools
4 toolscall_runai_apiCall Run:AI APIA
Run:AI REST API를 호출합니다. search_runai_api_spec으로 엔드포인트를 먼저 확인하세요. Bearer 토큰 인증은 자동. DELETE 등 위험 작업은 confirm_destructive: true 필요.
Inputs: path (required), method (required), payload, pathParams, +2 more.
execute_runai_cliExecute Run:AI CLIA
Run:AI CLI 전용 로컬 세션 작업 실행. API 대체 불가능한 작업만: port-forward (로컬 포트 → 워크로드 pod 터널링), exec (컨테이너 내 명령 실행: nvidia-smi, bash 등), attach (실행 중 프로세스에 TTY 연결), logs (로그 조회/스트리밍 --follow), diagnostics (클러스터 진단). 워크로드 생성/삭제/조회 등은 search_runai_api_spec + call_runai_api를 사용하세요.
Inputs: command_args (required), timeout_seconds.
resolve_runai_namesResolve Run:AI NamesA
Run:AI 엔티티 이름을 ID로 변환합니다. API는 ID를 요구하지만 사용자는 이름만 아는 경우가 대부분입니다. 지원: cluster, project, department, nodepool, workload, user, service-account, access-key. 여러 엔티티를 한번에 resolve 가능.
Inputs: resolve (required).
search_runai_api_specSearch Run:AI API SpecA
NVIDIA Run:AI OpenAPI 스펙을 검색하여 특정 작업에 필요한 API 엔드포인트, HTTP 메서드, 요청 스키마를 찾습니다. Run:AI의 모든 API를 커버합니다: 워크로드(workloads): training, inference, workspace, distributed(MPI/PyTorch/JAX/TensorFlow/XGBoost) 생성/조회/삭제/중지(suspend)/재개(resume), pod 목록, 이벤트, 메트릭 | 클러스터(clusters): 생성/조회/삭제, GPU 메트릭, 설치 정보, 스토리지 클래스 | 노드(nodes)/노드풀(node-pools): GPU/CPU 리소스 현황, 텔레메트리, 메트릭 | 조직: 프로젝트(projects), 부서(departments), 리소스 할당(quota), 우선순위, 리포트 | 인증/권한(RBAC): 역할(roles), 접근 규칙(access-rules), 서비스 계정, 액세스 키, 퍼미션 | 에셋(assets): 컴퓨트, 크레덴셜(docker-registry/NGC/S3/git/password/generic-secret), 데이터소스(config-map/host-path/NFS/PVC), 환경(environment), 레지스트리 | 정책(policies): 워크로드 타입별 리소스 제한/기본값 | 템플릿(templates): training/inference/workspace/distributed 재사용 템플릿 | 알림: Slack, 이메일 채널/구독 | 데이터볼륨(datavolumes), AI 애플리케이션 | 사용자, 감사 로그(audit), 테넌트 설정, 보안 설정, IDP. Deprecated 엔드포인트는 기본적으로 숨겨지며, 대체 API를 안내합니다.
Inputs: query (required), include_deprecated.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 4 tools
Each tool has a distinct responsibility: searching the API spec, executing API calls, resolving names to IDs, and running CLI-only operations. There is no overlap; dependencies like 'search before call' are explicitly documented.
All tool names follow a consistent verb_runai_noun pattern with snake_case: search_runai_api_spec, call_runai_api, resolve_runai_names, execute_runai_cli. The naming is uniform and predictable.
With only 4 tools, the server covers an enormous API surface by using a meta-design (search + call + resolve + CLI). Each tool is essential and well-scoped, avoiding redundancy.
The toolset provides comprehensive coverage: search_runai_api_spec exposes every API endpoint, call_runai_api executes them, resolve_runai_names bridges user-friendly names to API IDs, and execute_runai_cli handles operations impossible via API (port-forward, exec, logs). No obvious gaps.
Maintenance
Related MCP Connectors
214 tools for AI agents over one MCP endpoint: image, video, audio, 3D, data, storage, mail, GPUs.
MCP-Native LLM Orchestration Agent
The OpenRouter for tools. One MCP connection gives any AI agent 254 hosted tools, pay per call.
Free public MCP for AI agents — 193 tools, 44 workflows. No API key.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to access a unified catalog of tools from various APIs (OpenAPI, GraphQL, MCP, Google Discovery) through the MCP protocol.12 npmMIT
- AlicenseBqualityCmaintenanceEnables AI agents to consume 50+ NVIDIA NIM models including LLMs, image generation, and vision through 10 MCP tools with rich metadata and advanced filtering.11226 npm1MIT
- AlicenseNot gradedqualityAmaintenanceEnables LLMs to orchestrate AI agents and tasks, inspect DAG workflows, monitor cluster metrics, and perform safe database queries and statements through standardized MCP tools.189Apache 2.0
- AlicenseNot gradedqualityCmaintenanceEnables LLM agents to connect to multi-cloud DevOps control planes for Terraform plan inspection, drift detection and remediation, cost estimation, incident triage, and cloud resource discovery across Azure, AWS, and GCP. It supports autonomous workflows for PR review, FinOps optimization, and self-healing operations through MCP tools.MIT