Flexera’s MCP Server for Workload Optimization: Query Ocean and Elastigroup from Your AI Assistant

When you ask an AI assistant about cluster spend or rightsizing, Ocean and Elastigroup data can be a part of those conversations. The MCP server brings insights about your running environments into the AI tools you already use.  

Ask your AI assistant about nodes, logs, rightsizing  and spend 

Platform engineers, SREs and FinOps teams now do more of their work by talking to an AI assistant in the IDE. Claude, GitHub Copilot and Cursor have become the place where people reason about their systems. 

Until now, your Ocean and Elastigroup data wasn’t available there. To answer “what is this cluster costing me?” or “Which workloads can I rightsize?”, you had to leave your flow. That meant opening the  console, or writing scripts against the API and rewriting them for every new question. This context switching comes at a cost, and those add up over a day. 

What is Flexera’s MCP Server for Workload Optimization? Flexera’s MCP server for Workload Optimization connects your AI assistant directly to Ocean and Elastigroup. It’s built on the Model Context Protocol (MCP), a widely adopted open standard for connecting AI applications to external data and tools.  

How the MCP server works 

It’s a hosted service, so there’s nothing to install or run. You give your assistant one URL and a Spot API token generated in the console. The server identifies your organization, accounts and clusters from the token, so you never enter an account or cluster ID. 

Through it, your assistant can read: 

• Cost: what your Ocean clusters and Elastigroups are spending 

• Rightsizing opportunities: potential savings from tuning workload requests 

• Achieved savings: what rightsizing has already saved you 

• Inventory: your accounts, Ocean clusters, nodes, Elastigroups and cluster logs
  

Example prompts: what you can ask the MCP server 

If you can phrase it as a question, your assistant can look it up. Some suggested prompts to get you started: 

• “What is my total cost spend this month, broken down by account?” 

• “Where are my biggest rightsizing opportunities across all clusters?” 

• “How much has rightsizing actually saved us on the production cluster?” 

• “Show the nodes in cluster X and any recent errors in its logs.” 

• “List my Elastigroups and the cost breakdown for the most expensive one.” 

Answers come from live Ocean and Elastigroup data. Your assistant can also combine them with context it already has such as your Helm charts, your Terraform, or the pull requests you have open.  

See it in action 

How to connect the MCP server to Claude Code, VS Code, Cursor, or Claude Desktop  

Setup takes a few minutes.  

1. Get a Spot API token. Generate one in the console by following Create API tokens. The MCP server uses this token to perform ReadOnly actions alone.

2. Add the server to your MCP client. Pick your client: 

Claude Code: one command. 

claude mcp add --transport http spot https://console.spotinst.com/mcp \ 
  --header "Authorization: Bearer <your-spot-token>" 

 

VS Code (GitHub Copilot): one command. 

code --add-mcp 
'{"name":"spot","type":"http","url":"https://console.spotinst.com/mcp","headers":
{"Authorization":"Bearer <your-spot-token>"}}' 

 

Cursor: add this to ~/.cursor/mcp.json. 

{ 
  "mcpServers": { 
    "spot": { 
      "url": "https://console.spotinst.com/mcp", 
      "headers": { "Authorization": "Bearer <your-spot-token>" } 
    } 
  } 
} 

 

Claude Desktop: add this to claude_desktop_config.json (Settings → Developer → Edit Config), then restart the app. It uses the mcp-remote bridge, so you’ll need Node.js installed. 

{ 
  "mcpServers": { 
    "spot": { 
      "command": "npx", 
      "args": ["mcp-remote", "https://console.spotinst.com/mcp", 
               "--header", "Authorization: Bearer ${SPOT_TOKEN}"], 
      "env": { "SPOT_TOKEN": "<your-spot-token>" } 
    } 
  } 
} 

3. Now ask a question. Try “What are my Ocean clusters costing me?” 

That’s the whole setup. The server gets your account and cluster details from the token, so there’s nothing else to configure.  

MCP server security: permissions, storage, and encryption The server is built to be safe to plug into an AI assistant: 

• Your token, your permissions. Every request uses your own Spot API token, so the assistant can only see what you can see. Use a least-privilege token. 

• Nothing stored. The token is sent with each request. The server doesn’t keep credentials or data. 

• Read-only by default. The only write action allowed is attaching or detaching workloads on an existing rightsizing rule, and it needs an edit-capable token. The server blocks every other change before it reaches the Spot API. 

• Encrypted. All traffic runs over TLS. 

How the MCP server supports workload optimization 

The MCP server brings live cost, inventory, rightsizing opportunities and achieved savings directly into AI-assisted workflows. Unlike a standalone dashboard, it lets users explore this information conversationally and combine it with context such as Helm charts, Terraform configurations and pull requests, without changing how Ocean or Elastigroup optimizes workloads. 

Users can understand what is running, where savings opportunities remain and what optimization has already achieved, without switching tools or building custom API scripts. 

Try it today 

Flexera’s MCP server for Workload Optimization is available now to all Ocean and Elastigroup customers. Connect it at https://console.spotinst.com/mcp, ask your first question, and tell us what you’d like it to do next. Your feedback shapes the roadmap.