CloudYali MCP Server
The CloudYali MCP server connects your CloudYali account to AI assistants that speak the Model Context Protocol. Ask about your cloud bill in plain English and the assistant fetches the answer from your account — cost, budgets, savings recommendations, anomalies and inventory across AWS, Azure and GCP.
It is read-only. It sees exactly what your CloudYali login sees, and it cannot change anything.
What you can ask the CloudYali MCP server
- “Which five services cost us the most this month, across every cloud?”
- “How are we tracking against our budgets?”
- “Any cost anomalies in the last seven days, and what caused them?”
- “How much would we save if we actioned every open recommendation?”
- “Find every resource tagged environment=prod that is still running.”
- “What did instance i-0abc123 cost last month?”
Building reports with the CloudYali MCP server
Once a question is answered the assistant is holding the rows, so it can lay them out. Ask for the artefact rather than the figure:
- “Pull last month’s spend, budgets, anomalies and open savings, and build me a review for the leadership team.”
- “Show budget burn as of this morning, worst first, and flag anything over 90%.”
- “Chart daily spend for this account around the 29th, mark the anomaly, and write two paragraphs on what changed.”
- “Group open recommendations by how hard they are to action, not by how much they save.”
- “Same breakdown, this quarter against last, and only show what moved more than 15%.”
Worth setting as a standing rule with whichever assistant you use: any caveat that came with the data belongs on the page, not in the chat above it. The chart is what gets shared; the conversation stays behind.
Installing the CloudYali MCP server
The server is not on npm yet, so there is no npm install to give you today. Build it from source — it takes about a minute:
git clone https://github.com/cloudyali/cloudyali-mcp-server.git cd cloudyali-mcp-server npm install npm run build # then register it, e.g. with Claude Code claude mcp add cy -- node /path/to/cloudyali-mcp-server/dist/index.js
Requires Node.js 20 or newer. The source is MIT-licensed at cloudyali/cloudyali-mcp-server.
Config blocks for every client are in the MCP setup guide.
CloudYali MCP server client support
It is a standard local (stdio) MCP server, so it runs in any client that launches local MCP servers: Claude Code, Claude Desktop, Cursor, VS Code, OpenAI Codex CLI, Windsurf, Cline, Gemini CLI.
Browser-only assistants — ChatGPT on the web, claude.ai in a tab — cannot start a local process, so they cannot connect to this build. A hosted server is on the roadmap.
CloudYali MCP server actions
35 read-only actions across five areas. The assistant discovers the right one from your question, then runs it — you never name an action yourself.
| Area | Actions | Covers |
|---|---|---|
| Cost | 5 | Cost reports, spend aggregation, period comparison and the available filters. |
| Budgets | 7 | Budget list and detail, tracking summary, the resources driving spend, and spend, config and alert history. |
| Savings recommendations | 7 | Open recommendations, total identified savings, top opportunities, detail, history and assignees. |
| Cost anomalies | 4 | Anomaly list and summary, root cause detail, and the current alert configuration. |
| Asset inventory | 12 | Resource list, search and detail, counts, per-resource cost, change history, and the provider, type, region, account and tag lookups. |
Every action listed is a read. The full reference is in the MCP action reference.
CloudYali MCP server security
- Read-only. It cannot create, edit or delete anything. Budgets, tags and recommendation status are changed in the CloudYali portal, by a person.
- Your permissions, exactly. The server acts as you and inherits your scope. If you cannot see an account in the console, the assistant cannot see it either.
- Your sign-in. Authentication runs through the CloudYali console in your browser, behind your usual login including MFA and SSO. A verification code is shown on both sides — approve only if they match.
- Credentials stay local. The credential is written to your own home directory, readable only by your user, and refreshes itself. Delete the file to revoke that machine.
Full detail in the MCP security and permissions guide.
CloudYali MCP server support
Issues and pull requests are welcome at cloudyali/cloudyali-mcp-server, or mail support@cloudyali.io. An answer that looks wrong is worth reporting — include the question you asked.