Skip to main content

Spotto MCP: Connect AI to Azure Cost Data

Overview​

The Spotto MCP server connects AI assistants directly to your Azure cloud cost and resource data. Instead of copying data into prompts or switching between tools, you can ask questions about your cloud spend in natural language and get answers backed by real-time data.

What is MCP?​

The Model Context Protocol (MCP) is an open standard that allows AI assistants to securely connect to external data sources. Think of it as a bridge between your AI tools and your cloud data—the AI can query Spotto directly while you maintain control over authentication and access.

Compatible clients include:

Endpoint Profiles​

Spotto publishes three MCP endpoint profiles in each region:

ProfileBest forWhat it includes
fullGeneral-purpose clients that can load the whole Spotto catalogRead and write tools together
readAnalysis-heavy workflows and clients with tool-catalog limitsRead-oriented tools only
writeAction-only workflowsWrite-oriented tools only

Start with full unless your client struggles to retrieve the tool list. If it does, switch to the regional read endpoint instead. For the full matrix and selection guidance, see MCP Endpoint Profiles.

What Can You Do?​

Once connected, you can have natural conversations with your AI assistant about your cloud environment:

Cost Analysis​

Ask questions about your spending patterns, identify anomalies, and understand where your money goes.

"What's driving the cost increase in my production subscription this month?"

"Show me a breakdown of storage costs across all resource groups"

"Which resources have the highest spend this week?"

Resource Management​

Browse your Azure inventory, get detailed resource information, and track configuration changes.

"List all VMs in the analytics resource group with their current costs"

"What's the configuration of my SQL database server-prod-01?"

"Are there any Azure service retirements I need to plan for?"

Recommendations​

Access cost-saving opportunities, security improvements, and optimization suggestions—then act on them.

"What are my top 5 cost-saving recommendations?"

"Dismiss the VM rightsizing recommendation for batch - server—we need the capacity for month-end processing"

"Share the storage optimization recommendations with the infrastructure team on Slack"

Security Posture​

Monitor your secure score and review security recommendations.

"What's my current secure score across all subscriptions?"

"Show me high-priority security recommendations"

Regional Endpoints​

Spotto MCP is available in three regions. Each region has full, read, and write variants:

RegionDefault full endpoint
Australiahttps://mcp-au.spotto.ai/mcp
United Stateshttps://mcp-us.spotto.ai/mcp
Europehttps://mcp-eu.spotto.ai/mcp

If you need the matching -read or -write URL for your region, use the table in MCP Endpoint Profiles.

Data Isolation

Each region is completely separate. Use the endpoint that corresponds to where your Spotto account was created—there's no cross-region data access.

Troubleshooting​

Tools aren’t showing up in my client​

What you're seeing: The MCP server connects but your client shows no tools (or none from Spotto). Likely causes:

  • The MCP server URL is wrong for your region.
  • The API key is missing/invalid.
  • The endpoint profile is wrong for the client, or the client cannot load the full catalog. How to fix:
  1. Confirm you’re using the correct regional endpoint under Regional Endpoints above.
  2. If your client struggles to retrieve tools, switch from the default full endpoint to the regional read endpoint described in MCP Endpoint Profiles.
  3. Verify the client is sending X-API-Key: <YOUR_API_KEY> on every request.
  4. Re-run the client setup steps in Client Setup.

I’m seeing tool-limit or retrieval errors​

What you're seeing: The client reports errors such as "limiting number of tool calls to 70" or "couldn't retrieve the requested items". Likely causes:

  • The client cannot handle the full Spotto tool catalog on one MCP server. How to fix:
  1. Change the MCP URL to the matching regional read endpoint.
  2. Restart the client or reconnect the MCP server.
  3. Test with a simple prompt such as "Check the Spotto API status".

I’m getting authentication failures​

What you're seeing: Requests fail with 401/403 errors. Likely causes: API key is revoked/rotated, or restricted by IP/country. How to fix:

  1. Rotate the key in the Spotto Portal and update your client configuration.
  2. If you enabled IP/country restrictions, confirm your client’s outbound IP matches the allowlist.

How It Works​

  1. Generate an API key from the Spotto Portal (see API Authentication)
  2. Configure your AI client with the Spotto MCP endpoint and your API key
  3. Start asking questions about your cloud environment

The AI assistant sends your questions to Spotto's MCP server, which queries your cloud data and returns the results. All communication is encrypted and authenticated with your API key. For more details on the underlying API, see the Spotto API documentation.

Security​

  • Authentication: Every request requires a valid Spotto API key
  • IP Allowlisting: Optionally restrict API key usage to specific IP addresses
  • Role-Based Access: Your API key inherits the permissions of the user who created it
  • Encryption: All traffic uses HTTPS
  • Edge Network: The MCP server runs on Cloudflare's global edge network for low latency

Next Steps​