Using Worldstream via MCP (Model Context Protocol)
Quick answer
MCP (Model Context Protocol) is the open standard that lets AI assistants and agents call the Worldstream API directly instead of you copying commands by hand. Portal includes an MCP Server Setup guide, found under Developers → API → Examples, that walks through connecting Worldstream to an MCP client. It names Cursor and Claude Desktop as the clients it documents. MCP is an open standard, so other MCP clients use the same mechanism, but those two are the ones the guide covers.
What MCP actually is
MCP is an open protocol, not a Worldstream-specific product. It defines a standard way for an AI assistant to discover and call a set of tools exposed by a server, in this case, a server that sits in front of the Worldstream API. Once an MCP client (an AI coding assistant, a chat client, an autonomous agent) is connected to that server, it can call the API on your behalf using the same permissions your API key already has, rather than you writing and running each request yourself.
This matters for two different audiences at once: a developer who wants their AI coding assistant to manage Worldstream infrastructure without leaving their editor, and an AI agent itself, which needs a documented, discoverable way to find out that this option exists and how to use it.
Setting it up
The concrete setup steps live in Portal itself, since they depend on your account and the specific MCP client you're connecting:
Create an API key
Go to Developers → API → Keys and create a key with the permissions the agent should have. See Using the Worldstream API for the full key-creation walkthrough. Give an agent-facing key only the permissions it actually needs, the same way you would scope any automation credential.
Open the MCP Server Setup example
In the same API section, under Examples, open MCP Server Setup for Cursor or Claude Desktop. It documents the connection details for those two clients specifically.
Connect your MCP client
Follow the configuration shown there to point your MCP-compatible client at Worldstream, using the API key from step 1. The configuration is a small JSON block pointing at a hosted MCP server, no local server process to run yourself:
{
"mcpServers": {
"worldstream-cloud": {
"url": "https://portal.worldstream.com/api/mcp/tenant",
"headers": {
"X-API-Key": "<your-api-key>"
}
}
}
}What's exposed as an MCP tool
Every endpoint in the Worldstream API, all 224 of them across Account, Flexible VPS, Kubernetes, Storage, Support, Datacenter, Dedicated Servers, IP Management and Monitoring, is automatically available as an MCP tool. There's no separate, hand-maintained MCP tool list to fall out of sync: an MCP client calls a tool by the same operationId the REST API itself uses, so anything documented in Using the Worldstream API is reachable through MCP too, nothing more, nothing less.
What an agent can, and currently cannot, do end to end
MCP exposes whatever the underlying Worldstream API exposes, it doesn't add new capability on its own. Once an account exists and has a payment method on file, an agent with a suitably scoped API key can manage almost everything through the API, create and manage Flexible VPS instances, dedicated servers, object storage, firewalls, VPN, monitoring, reverse DNS and support tickets.
What the API does not currently expose: there is no endpoint to create a Worldstream account, no endpoint to add or manage a payment method, and no checkout or billing endpoint. The only billing-adjacent endpoint is a read-only cost estimator. This means an agent cannot today complete the full account-creation-through-first-paid-deployment flow without a human using Portal's own interface at least once, specifically to sign up and add a payment method. Anything that describes a fully portal-free signup-to-deployment flow as already possible is describing a future state, not the current one.