Novita MCP Server
OfficialThe Novita MCP Server allows users to manage GPU instances and related resources on the Novita AI platform. With this server, you can:
List clusters/regions, products, GPU instances, templates, and network storage
Manage GPU instances: Create, start, stop, delete, restart, and retrieve details
Manage templates: Create, retrieve, and delete templates for GPU instances or serverless applications
Manage network storage: Create, list, update, and delete network storage resources
Manage container registry authentication: List, create, and delete authentication credentials
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., "@Novita MCP Server list my running GPU instances"
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.
Novita MCP Server
novita-mcp-server is a Model Context Protocol (MCP) server that provides seamless interaction with Novita AI platform resources. We recommend accessing this server through Claude Desktop, Cursor, or any other compatible MCP client.
Features
⚠️ Beta Notice:
novita-mcp-serveris currently in beta and only supports GPU instance management. Additional resource types will be supported in future releases.
Currently, novita-mcp-server enables management the resources of GPU instances product.
Supported operations are as follows:
Cluster(/Region): List;
Product: List;
GPU Instance: List, Get, Create, Start, Stop, Delete, Restart;
Template: List, Get, Create, Delete;
Container Registry Auth: List, Create, Delete;
Network Storage: List, Create, Update, Delete;
Related MCP server: Outsource MCP
Installation
You can install the package using npm, or Smithery:
Using npm
npm install -g @novitalabs/novita-mcp-serverUsing Smithery
Visit the https://smithery.ai/server/@novitalabs/novita-mcp-server and follow the "Install" instructions to install the server.
Configuration to use novita-mcp-server
First, you need to get your Novita API key from the Novita AI Key Management.
And next, you can use the following configuration for both Claude Desktop and Cursor:
📌 Tips
For Claude Desktop, you can refer to the Claude Desktop MCP Quickstart guide to learn how to configure the MCP server.
For Cursor, you can refer to the Cursor MCP Quickstart guide to learn how to configure the MCP server.
{
"mcpServers": {
"@novitalabs/novita-mcp-server": {
"command": "npx",
"args": ["-y", "@novitalabs/novita-mcp-server"],
"env": {
"NOVITA_API_KEY": "your_api_key_here"
}
}
}
}Examples
Here are some examples of how to use the novita-mcp-server to manage your resources with Claude Desktop or Cursor:
List clusters
List all the Novita clustersList products
List all available Novita GPU instance productsList GPU instances
List all my running Novita GPU instancesCreate a new GPU instance
Create a new Novita GPU instance:
Name: test-novita-mcp-server-01
Product: any available product
GPU Number: 1
Image: A standard public PyTorch/CUDA image
Container Disk: 60GBTesting
This project uses Jest for testing. The tests are located in the src/tests directory.
You can run the tests using one of the following commands:
npm testAvailable Tools
20 toolscreate-container-registry-authD
Inputs: name (required), username (required), password (required).
create-gpu-instanceD
Inputs: name (required), productId (required), gpuNum (required), rootfsSize (required), +7 more.
create-network-storageD
Inputs: clusterId (required), storageName (required), storageSize (required).
create-templateD
Inputs: template (required).
delete-container-registry-authD
Inputs: id (required).
delete-gpu-instanceD
Inputs: instanceId (required).
delete-network-storageD
Inputs: storageId (required).
delete-templateD
Inputs: templateId (required).
get-gpu-instanceD
Inputs: instanceId (required).
get-templateD
Inputs: templateId (required).
list-clustersD
Inputs: no named parameters.
list-container-registry-authsD
Inputs: no named parameters.
list-gpu-instancesD
Inputs: pageSize, pageNumber, name, productName, +1 more.
list-network-storageD
Inputs: pageNo, pageSize, storageName, storageId.
list-productsD
Inputs: clusterId, productName.
list-templatesD
Inputs: pageSize, pageNum, type, channel.
restart-gpu-instanceD
Inputs: instanceId (required).
start-gpu-instanceD
Inputs: instanceId (required).
stop-gpu-instanceD
Inputs: instanceId (required).
update-network-storageD
Inputs: storageId (required), storageSize (required), storageName.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 20 tools
Each tool has a clearly distinct purpose targeting specific resources and actions. The naming convention makes it easy to differentiate between operations on different resource types like GPU instances, templates, network storage, and container registry auths. There is no ambiguity about which tool to use for a given task.
Tool names follow a perfectly consistent verb_noun pattern throughout, using snake_case uniformly. The pattern is clear: action_resource (e.g., create_gpu_instance, list_templates, delete_network_storage). This consistency makes the tool set highly predictable and easy to navigate.
With 20 tools, the count is slightly high but reasonable for a cloud infrastructure server covering multiple resource types. It provides comprehensive operations for GPU instances, templates, network storage, and container registry auths, though it might feel a bit heavy compared to more focused servers.
The tool set offers good CRUD/lifecycle coverage for its domain, with create, delete, list, and specific operations like start/stop/restart for GPU instances. Minor gaps exist, such as missing update operations for GPU instances and templates, but agents can work around these with the available tools.
Maintenance
Related MCP Connectors
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
Hosted Notion MCP server for Claude, ChatGPT and Cursor, with each action switched per account.
Hosted OpenAI Ads MCP server for Claude, ChatGPT and Cursor, with each action switched per account.
One MCP server for the services your AI needs: memory, files, Microsoft 365, kanban and more.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceA proxy server that unifies multiple MCP servers, enabling seamless tool, prompt, and resource management via the MetaMCP App.124 npm154Apache 2.0
- AlicenseAqualityDmaintenanceAn MCP server that enables AI applications to access 20+ model providers (including OpenAI, Anthropic, Google) through a unified interface for text and image generation.230MIT
- AlicenseAqualityCmaintenanceMCP server for AI-powered image, audio, and video generation, enabling media creation directly from Claude, Cursor, and other MCP clients.1141 npmMIT

RunComfy MCPofficial
AlicenseAqualityCmaintenanceOfficial remote MCP server for RunComfy. 31 tools across three products: Serverless API (ComfyUI) deployment management and async inference, Model API hosted-model inference, and Trainer API LoRA training jobs.31MIT