Set Up AI Hub
Connect AI Hub to your CloudZero account and start asking questions about your costs. CloudZero hosts the MCP server, so you always connect to the current version. New MCP server tools and fixes reach you without a reinstall. Choose the setup path that fits your AI tool:
| Path | Best for | What it does |
|---|---|---|
| Option A: Claude Code Plugin | Claude Code users | Connects to the MCP server and installs guided analysis skills in one command |
| Option B: Other tools quick setup | Cursor, VS Code, Codex, and other supported tools | Connects to the MCP server with a single npx command |
| Option C: Manual configuration | MCP-compatible AI tools | Configures the MCP server connection directly |
Connecting an enterprise agent platform such as Amazon Bedrock AgentCore or Google Gemini Enterprise takes a different path, because those platforms need credentials issued in advance. See Connect an Agent Platform with MCP OAuth Clients.
Option A: Connect with the CloudZero Plugin for Claude Code
The fastest way to get started. The Claude Code Plugin connects you to the CloudZero MCP server and installs guided analysis skills in one command.
What you need
- A CloudZero account
- Claude Code (v2.1 or above). If you do not have Claude Code installed, visit Anthropic's Claude Code installation guide, select your operating system and preferred install method, and follow the steps. Once
claudeis available in your terminal, return here to continue. - (Recommended) A dedicated folder for your cost analysis work (for example,
cloudzero-cost-analysis). Navigate to this folder in your terminal before startingclaude.
The CloudZero Plugin is supported through the Claude Code CLI. This is separate from claude.ai and Claude Desktop.
Install
- Run the following commands to install the plugin:
claude plugin marketplace add cloudzero/cloudzero-claude-marketplace claude plugin install cost-analyst@cloudzero
The CloudZero marketplace also includes Model Right Sizer for optimizing model selection across tasks.
-
Start a Claude Code session by running
claudein your terminal. -
Run
/mcpand selectplugin:cost-analyst:cloudzerofrom the list. -
Follow the instructions to sign in to CloudZero.
-
After signing in to CloudZero, run the following command to generate organization context from your recent query history: what your Dimensions mean, common analysis workflows, and cost allocation patterns. This improves the accuracy and relevance of every query. Run it again any time to refresh the generated context. It does not change the Custom Instructions you have added, so rules you give the agent persist across runs. For details, see Context tools in the MCP Server Reference.
/plugin:cost-analyst:cloudzero:init -
You are ready to start querying. Try one of the examples below.
Connect
To start a new session:
- Navigate to your cost analysis folder and run
claudein your terminal. - Run
/mcpand selectplugin:cost-analyst:cloudzerofrom the list. - Sign in to CloudZero when prompted. If you already have an active session, authentication is automatic.
Option B: Quick setup for other MCP-compatible AI tools
Connect any supported AI tool to the CloudZero MCP server with a single command using npx add-mcp.
What you need
- A CloudZero account
- Node.js (v18 or later)
Add with npx
Run the add command for your AI tool:
| AI tool | Add command |
|---|---|
| Antigravity | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a antigravity -y |
| Claude Code | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a claude-code -y |
| Cline (VS Code) | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a cline -y |
| Cline CLI | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a cline-cli -y |
| Codex | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a codex -y |
| Cursor | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a cursor -y |
| Gemini CLI | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a gemini-cli -y |
| GitHub Copilot CLI | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a github-copilot-cli -y |
| Goose | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a goose -y |
| MCPorter | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a mcporter -y |
| OpenCode | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a opencode -y |
| VS Code | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a vscode -y |
| Zed | npx add-mcp https://czca-server.discovery.cloudzero.com/mcp -g -a zed -y |
These commands register the hosted endpoint in the tool's global configuration. To register it in a project-level configuration instead, remove the
-gflag. For the most updated list of supported tools and commands, see the npx add-mcp documentation.
Install for Kiro
Kiro uses its own install command. For details, see the CloudZero Kiro Power repository.
kiro /powers install cloudzeroConnect
Open your AI tool and invoke any CloudZero tool. On first use, a browser window opens to sign in to CloudZero. If you already have an active session, authentication is automatic.
Option C: Manually connect an AI tool via MCP
Use this path if your AI tool is not listed above or you want full control over your MCP configuration.
Configure
Add the CloudZero MCP server to your AI tool's configuration using the following values:
| Setting | Value |
|---|---|
| Server name | cloudzero |
| URL | https://czca-server.discovery.cloudzero.com/mcp |
| Type | http |
| Authentication | OAuth (a browser window opens on first use). For API key authentication, see API Key Authentication. For an enterprise agent platform that needs a client ID and secret, see Connect an Agent Platform with MCP OAuth Clients. |
Most AI tools accept this as a JSON configuration block:
{
"mcpServers": {
"cloudzero": {
"type": "http",
"url": "https://czca-server.discovery.cloudzero.com/mcp"
}
}
}Refer to your AI tool's documentation for where to add MCP server configurations.
Connect
On first use with OAuth, a browser window opens to sign in to CloudZero. If you already have an active session, authentication is automatic.
What to expect
After setup, you can start querying your cost data immediately. AI Hub accesses your full CloudZero data model, including the custom business logic in your Dimensions, optimization recommendations, and data from every connected provider. The same spend intelligence your team has built in CloudZero is available through natural language.
Examples of prompts you can ask
- "We had a cost spike last Tuesday. Investigate what caused it and recommend actions"
- "Do a deep dive on our RDS costs and identify optimization opportunities"
- "Generate a cost breakdown by team for Q4 including service details"
- "Analyze our tag coverage and prioritize resources that need tagging"
- "Run an anomaly detection scan, then show me my top cost drivers and any trends"
For more prompts organized by role and use case, see the Prompt Catalog.
Have questions or feedback? Reach out to your account manager.
Updated 14 days ago

