Model Context Protocol (MCP) — Beta

Connect your AI tools to Substrate Enterprise using MCP

---

The Model Context Protocol (MCP) is a standard for connecting Large Language Models (LLMs) to platforms like Substrate Enterprise. Once connected, your AI assistants can interact with and manage your enterprise cloud infrastructure on your behalf—provisioning VMs, managing storage, configuring networks, load balancers, and more.

Beta Feature

The Substrate Enterprise MCP server is currently in beta. We welcome feedback as we refine the experience.

Connect to the MCP Server

Step 1: Configure your MCP client

Add the Substrate Enterprise MCP server to your MCP client configuration. You can do this in one of two ways:

Option A: Add configuration manually

1. Open or create your MCP client's configuration file (e.g. mcp.json in your project root or config directory)
2. Add the following configuration:

{
  "mcpServers": {
    "cloud-compute-storage": {
      "url": "https://mcp.substrateai.net/mcp",
      "headers": {
        "x-api-key": "YOUR_API_KEY"
      }
    }
  }
}

Replace YOUR_API_KEY with your Substrate Enterprise API key. You can obtain this from the Enterprise Dashboard under Settings > API Keys.

Option B: Use environment variable for authentication

If you prefer to keep your API key out of the config file:

{
  "mcpServers": {
    "cloud-compute-storage": {
      "url": "https://mcp.substrateai.net/mcp"
    }
  }
}

Then set the CLOUD_API_KEY environment variable before launching your MCP client:

export CLOUD_API_KEY="your-api-key"

Step 2: Restart your MCP client

After adding the configuration, restart your MCP client to load the MCP server. The server will be detected automatically.

Step 3: Verify the connection

1. Navigate to your client's settings or tools panel
2. Confirm that cloud-compute-storage appears in the list of connected MCP servers
3. Test by asking a natural language question, for example: "What regions are available?" or "List my VMs." — your AI assistant will use the MCP tools to respond

---

Agent Context Document

For AI agents and developers integrating with the MCP server, we provide an Agent Context document that summarizes tool usage, workflows, and best practices.

Download Agent Context

The Agent Context document includes:

• Tools at a glance — Quick reference for all 10 tools and 85 actions
• Workflow recipes — Full-stack deploy, microservices, VM creation, storage, and more
• User intent → tool mapping — How to translate natural language requests into tool calls
• Safety rules — Best practices for read-before-write, stop-before-resize, and quota awareness

---

Available Tools

The Substrate Enterprise MCP server provides 10 workflow-optimised tools across compute, storage, networking, load balancing, and security:

Tool	Actions	Domain
discover_infrastructure	2	Projects, regions, quotas
provision_vm	—	Quota-aware VM creation
provision_baremetal	—	Quota-aware bare metal creation
provision_load_balancer	—	LB creation with quota pre-check
manage_compute	17	VM lifecycle, resize, interfaces, catalog
manage_storage	11	Volumes, snapshots, attach/detach/extend
manage_networking	16	Networks, subnets, routers
manage_ips	12	Floating IPs, reserved fixed IPs
manage_security	10	Security groups, SSH keys
manage_load_balancers	14	LB lifecycle, listeners, pools, members

For the complete parameter-level reference, see MCP Capabilities.

---

Authentication

Authentication is supported via:

• API key header: x-api-key
• Environment variable: CLOUD_API_KEY

Generate your API key from the Enterprise Dashboard under Settings > API Keys.

---

Next Steps

• MCP Capabilities — Full technical reference for every tool and parameter
• Virtual Instances — Guide to launching and managing VMs
• Bare Metal — Deploy dedicated GPU servers

---

Need help? Contact enterprise@substrate.ai for support.