Everyone is talking about MCP for code. Autocomplete, refactoring, chat-driven development. Almost nobody is using MCP to manage cloud infrastructure.
MCP (Model Context Protocol) is an open standard that lets AI tools like Cursor and Claude connect to external systems and execute real actions — not just answer questions.
We built this at Quave ONE: 62 MCP tools that give your AI assistant direct access to your infrastructure. Not a chatbot. Actual operational control — deploy, scale, roll back, configure alerts, inspect logs. All from your editor, in plain English.
We were one of the first cloud platforms to build this.
What it looks like in practice
"How's production looking this morning?"
→ CPU at 42%, memory at 68%, no error spikes overnight. 3 deploys yesterday, all healthy."Response times are slow on the API. What changed?"
→ Deploy at 14:15 correlates with the latency spike. P95 went from 120ms to 2.4s. Recommend rolling back."Roll back the last deploy."
→ Rollback complete. Health checks passing. P95 back to 130ms."Scale the worker to 4 instances."
→ Done. Zero downtime. All instances healthy."Set a CPU alert at 75% for 5 minutes. Send to Slack."
→ Alert configured.
Real operations. Production infrastructure. Controlled through conversation.
Watch it in action: youtube.com/watch?v=rFOPE-Kqxmg
Why this matters for small teams
If you have 8 engineers, your "platform team" is whoever touched Terraform last. Your runbook is a Slack thread from 6 months ago.
MCP changes this. Your AI coding assistant becomes your ops assistant.
CerradoX debugged a deploy failure without opening a single tool. From Cursor. In plain English. "The speed this enables for interaction with agents is massive."
They don't have a DevOps engineer. They don't need one.
What the 62 tools cover
- Deployments: deploy, roll back, view history, compare versions
- Scaling: scale up/down, check resource usage
- Logs: view, filter, search by pattern
- Metrics: CPU, memory, request rate, error rate, latency
- Alerts: create and configure with real threshold data
- Environment variables: view, set, and update across environments
- Apps and databases: create, configure, check backup status
- Domains: manage custom domains and SSL
All secured with server-enforced scopes and encrypted keys.
Setup: 3 steps
- Deploy your app on Quave ONE (or migrate — free)
- Generate an MCP API key
- Add the server URL to Cursor, Claude Desktop, or Claude Code
No SDK. No integration code. Full guide: docs.quave.cloud/mcp
$7.50/month per zcloud. MCP included.
Everything included: hosting, databases, deployments, observability, alerts, and 62 MCP tools. No premium tier. No add-on.
No credit card required. Start here: docs.quave.cloud/mcp
Frequently asked questions
What is MCP (Model Context Protocol)?
An open standard that lets AI tools like Claude and Cursor connect to external systems and execute real actions — not just answer questions.
Can I manage cloud infrastructure with MCP?
Yes. Quave ONE provides 62 MCP tools for deploying, scaling, rolling back, viewing logs, checking metrics, configuring alerts, and managing your entire infrastructure through natural language.
Is MCP safe for production?
Quave ONE uses server-enforced scopes and encrypted API keys. Your AI can only perform actions within the permissions you configure. All operations are logged.
How is MCP different from a chatbot?
Chatbots answer questions from docs. MCP executes real operations on your production infrastructure.
Does MCP replace DevOps engineers?
For teams under 15 engineers, MCP eliminates most of the operational work that would require a dedicated DevOps hire. It doesn't replace deep expertise — it makes routine operations instant.