Hyperstack Docs MCP Server
Connect MCP-compatible AI tools directly to the Hyperstack documentation for grounded answers with citation links.
The Hyperstack Docs MCP Server searches the documentation when you ask a question in your AI client, and returns grounded answers with source links. Responses reflect the current documentation instead of a model's training data. The server is read-only. It cannot create, change, or delete any Hyperstack resource.
The Hyperstack API MCP Server is a different server. It creates and manages real resources (Virtual Machines, clusters, volumes, billing) through the Hyperstack API. You run it yourself and it requires an API key. Use this page's Docs MCP Server to look things up, and the API MCP Server to take action.
Capabilities
With the Docs MCP Server connected, your AI client can:
- Answer Hyperstack questions from the docs, with a link to each source page.
- Retrieve a full documentation page when it needs complete parameter tables or code examples.
- Browse the documentation structure by product area (Virtual Machines, Kubernetes, storage, networking, billing, AI Studio, hardware).
- Explain a Hyperstack API error by status code, error reason, or message text, and point to the fix.
- Look up GPU flavor specifications when comparing hardware or planning a deployment.
Connect to the server
The Docs MCP Server is hosted by Hyperstack at https://docs.hyperstack.cloud/mcp. There is nothing to install or run, and no API key is required. Any MCP-compatible client can connect over Streamable HTTP with no authentication.
Choose your client. One click gives you what it needs to connect.
The server URL, for any MCP client (Streamable HTTP, no authentication):
https://docs.hyperstack.cloud/mcp
You can also open the server URL in a web browser for the same connect options and the list of tools. The sections below give the command or configuration for each client.
Claude Code
Execute the following command:
claude mcp add --transport http hyperstack-docs https://docs.hyperstack.cloud/mcp
Claude Desktop
Add the Docs MCP Server as a custom connector:
- Navigate to Settings.
- Under Customize, click Connectors.
- Click Add, then click Add custom connector.
- Enter a Name for the server, such as
Hyperstack Docs. - Enter
https://docs.hyperstack.cloud/mcpas the URL. - Click Add.
The server requires no authentication, so leave Advanced settings empty.
You can also connect through mcp-remote instead. Add the following to your claude_desktop_config.json, then restart Claude Desktop:
{
"mcpServers": {
"hyperstack-docs": {
"command": "npx",
"args": ["mcp-remote", "https://docs.hyperstack.cloud/mcp"]
}
}
}
Codex
Execute the following command:
codex mcp add hyperstack-docs --url https://docs.hyperstack.cloud/mcp
ChatGPT
ChatGPT connects to custom MCP servers in Developer mode (Pro, Plus, Business, Enterprise, and Education plans on the web). Turn on Developer mode in ChatGPT settings, then add https://docs.hyperstack.cloud/mcp as a connector with No Authentication.
VS Code
Click VS Code above to install the server directly. To add it by hand, add the server to .vscode/mcp.json in your workspace (or run MCP: Add Server from the Command Palette and choose HTTP):
{
"servers": {
"hyperstack-docs": {
"type": "http",
"url": "https://docs.hyperstack.cloud/mcp"
}
}
}
Cursor
Click Cursor above to install the server directly. To add it by hand, add the server to your ~/.cursor/mcp.json file (or a project-level .cursor/mcp.json):
{
"mcpServers": {
"hyperstack-docs": {
"url": "https://docs.hyperstack.cloud/mcp"
}
}
}
Gemini CLI
Execute the following command:
gemini mcp add --transport http hyperstack-docs https://docs.hyperstack.cloud/mcp
Any agent: paste a setup prompt
If your AI client can edit its own configuration (Claude Code, Codex, Gemini CLI, Cursor, and other coding agents), paste this prompt into it. The agent adds the server, then runs a search to confirm the connection works.
Add the Hyperstack Docs MCP server to this agent: name hyperstack-docs, URL https://docs.hyperstack.cloud/mcp, Streamable HTTP, no auth. Setup guide: https://docs.hyperstack.cloud/docs/libraries/docs-mcp/. Then run its search_docs tool for "Which regions are available?" to confirm it works. If you cannot add MCP servers, use https://docs.hyperstack.cloud/llms.txt instead.
All installed agents at once
add-mcp detects the AI clients installed on your machine and adds the server to each of them:
npx add-mcp https://docs.hyperstack.cloud/mcp --name hyperstack-docs
Add -g to install for all projects instead of the current directory.
Verify the connection
Once the server is added, confirm it is working by asking your AI client a Hyperstack question, for example:
How do I create a spot Virtual Machine on Hyperstack?What GPU flavors are available with an H100?What does a Hyperstack 409 already_exist error mean?
If the connection is working, the client calls the Docs MCP Server and returns an answer that includes one or more https://docs.hyperstack.cloud/docs/... source links.
In Claude Code, you can also list your configured servers and check the connection status:
claude mcp list
Available tools
The server exposes five read-only tools. Your AI client calls them automatically based on your question. For the full parameter reference and example prompts, see the Tools Reference.
| Tool | Purpose |
|---|---|
search_docs | Search the documentation for a concept, feature, API endpoint, or workflow. |
get_page | Retrieve the full content of a specific documentation page. |
list_sections | List the documentation structure, optionally filtered by product area. |
lookup_error | Explain a Hyperstack API error and point to the fix. |
get_flavor_info | Return GPU flavor specifications for a model or hardware query. |
Related Documentation
- Tools Reference covers every tool, its parameters, and example prompts.
- Hyperstack API MCP Server creates and manages Hyperstack resources through natural language.