Connection Guide
The Hyperstack API MCP Server enables natural-language infrastructure management through MCP-compatible AI clients.
Instead of writing API calls by hand, you describe what you want in plain English, and the Hyperstack API MCP (Model Context Protocol) Server translates the request into an authenticated Hyperstack API operation.
This guide walks you through connecting an AI client to the Hyperstack-hosted MCP Server.
This MCP server can create, modify, and delete real Hyperstack resources, which may incur charges on your account. Always review actions before confirming them. NexGen Cloud is not liable for any unintended resource usage, costs, or data loss resulting from the use of this tool. Use it at your own risk.
About the Hyperstack API MCP Server
The MCP Server works with any MCP-compatible AI client, including Claude Code, Claude Desktop, and Open WebUI. It presents Hyperstack operations as tools your client can call, and turns each call into an authenticated request to the Hyperstack Infrahub API.
Once connected, you can:
- Create and manage Virtual Machines
- Provision and scale Kubernetes clusters
- Create and attach Volumes
- Retrieve Billing and usage information
- Manage Environments
- Execute multi-step infrastructure workflows
Connect to the hosted MCP Server
Hyperstack hosts an MCP Server, so you can connect an AI client without installing or running anything yourself. Point your client at this URL:
https://console.hyperstack.cloud/ai/mcp
The hosted server authenticates you through your Hyperstack account. The first time a client connects, your browser opens the sign-in page. Once you sign in, the client receives its own credentials and every tool call runs against your account, so there is no API key to configure.
Claude Code
Claude Code connects over HTTP, so it needs no proxy and no additional packages. Execute the following command to add the server:
claude mcp add --transport http hyperstack-mcp https://console.hyperstack.cloud/ai/mcp
The server is added at the local scope, which makes it available to you in the current project. Add --scope user to make it available across all your projects, or --scope project to share it with your team through a .mcp.json file.
Adding the server does not sign you in. To sign in, start a Claude Code session, run /mcp, select hyperstack-mcp, and complete the sign-in in the browser window that opens. Claude Code stores the resulting credentials and renews them for you, so you sign in once.
Execute the following command to confirm the connection:
claude mcp list
Until you have signed in, this command reports the server as needing authentication.
Claude Desktop
Claude Desktop connects to the hosted server as a custom connector, natively over HTTP, so it needs no proxy and no additional packages.
a. Open Claude Desktop and go to Settings
b. Select Connectors
c. Click Add, then Add custom connector
d. In the Name field, enter a name for the server, for example Hyperstack
e. In the Remote MCP server URL field, enter https://console.hyperstack.cloud/ai/mcp, then click Add
The connector appears in the Connectors list. Click Connect to sign in, and complete the sign-in in the browser window that opens. Claude Desktop stores the resulting credentials and renews them for you, so you sign in once.
To confirm the connection, start a new chat and ask Claude to show you all Hyperstack tools.
Connecting Claude Desktop to an instance you run yourself uses the mcp-remote proxy instead. For that procedure, including the configuration file location for each operating system, see the Claude Desktop Setup Guide.
Open WebUI
Open WebUI connects to the MCP Server as a tool server. If you do not already have Open WebUI running, start it with Docker:
docker run -d \
-p 3000:3000 \
-e PORT=3000 \
--name open-webui \
ghcr.io/open-webui/open-webui:main
Open WebUI is available at http://0.0.0.0:3000.
-
Configure the MCP Server in Open WebUI
a. Navigate to Admin Panel, switch to the Settings tab, then select Integrations under Tools
b. Click the + icon under External Tool Servers to add a new connection
c. Click the Type toggle to change it fromOpenAPItoMCP Streamable HTTP
d. Set the API Base URL tohttps://console.hyperstack.cloud/ai/mcp
e. Enter a Name. Leave ID asautounless you need a specific value
f. Set Auth toOAuth 2.1, so Open WebUI signs you in with your Hyperstack account
g. Click Register Client. The badge beside it changes from Not Registered to Registered
h. Click SaveRegister the client before you saveOpen WebUI refuses to save an MCP connection that uses
OAuth 2.1until you have registered the client, and the Authorize OAuth button stays disabled until the connection is saved. Complete these steps in order. -
Authorize Open WebUI
Reopen the connection you just saved and click Authorize OAuth. Your browser opens the Hyperstack sign-in page. Sign in and approve access, then return to Open WebUI. Open WebUI stores the resulting credentials and renews them for you, so you sign in once.
-
Add an AI Studio Model
a. Navigate to Admin Panel → Settings → Connections
b. Under OpenAI API, click + to add a new connection
c. Set the server URL tohttps://console.hyperstack.cloud/ai/api/v1
d. Add your Hyperstack API key underAuthorization: Bearer <YOUR_API_KEY>
e. (Optional) If you know the specific model you want to use, add it under Model IDs and click +
f. Click SaveAfter saving, the newly added model should be visible in the Chat section.
-
Configure Function Calling and Enable MCP Tools
Once both the model and MCP server have been set up, complete the following steps within a New Chat session.
a. Navigate to New Chat
b. Toggle Controls and set Function Calling toNative
c. Open Integrations → Tools and enable the MCP Server connection you configured -
Validate the Integration
If setup is successful, you should be able to:
- See the newly added model available in chat
- Enable the MCP tools for the session
- Run prompts such as:
List all virtual machines in my accountShow current billing informationWhat environments are available?
Open WebUI invokes the appropriate MCP tool through the Hyperstack API MCP Server and returns the results.
Next StepsFor a complete list of supported operations and additional example usage, see the full guide: MCP Tools and Operations Reference.
Run your own instance
To run the MCP Server inside your own network, or to authenticate it with a single API key rather than per user, see the Local Installation Guide. That guide covers when to choose a self-hosted instance, how to run it with Docker, and how to point the clients above at it.
Troubleshooting
If you encounter any issues:
- Confirm the MCP connection URL is set to
https://console.hyperstack.cloud/ai/mcp - Complete the sign-in in the browser window your client opens, and sign in with the Hyperstack account whose resources you want to manage
- Confirm your client appears as connected, using
claude mcp listin Claude Code or the tool server entry in Open WebUI
For detailed troubleshooting guidance, see the Troubleshooting Guide.
Related Documentation
- Claude Desktop Setup Guide: Add the hosted server as a custom connector, or reach your own instance through the
mcp-remoteproxy. - MCP Tools and Operations Reference: Every supported tool, grouped by domain, with example prompts.
- Self-hosted MCP Server: Run your own instance when the hosted server does not fit, with its installation and configuration guides.
- Troubleshooting Guide: Hosted sign-in problems, plus Docker, authentication, and client connectivity on your own instance.