Skip to content

CloudAEye in IntelliJ IDEA

Connect CloudAEye through JetBrains AI Assistant and run its skills with Claude Agent or Codex. This setup uses an MCP connection and manually imported skills, rather than a packaged JetBrains plugin.

1. Prepare AI Assistant

Install or enable JetBrains AI Assistant under Settings > Plugins. Update IntelliJ IDEA and the plugin until Settings > Tools > AI Assistant > Skills is available.

Install the required runtimes, including Node.js/npm for the OAuth bridge.

2. Connect MCP and authenticate

Open Settings > Tools > AI Assistant > Model Context Protocol (MCP). Click Add, choose STDIO, and use this Windows configuration:

{
  "mcpServers": {
    "cloudaeye": {
      "command": "npx.cmd",
      "args": [
        "-y",
        "--ignore-scripts",
        "mcp-remote@0.14.3",
        "https://api.cloudaeye.com/mcp",
        "--transport",
        "http-only",
        "--static-oauth-client-metadata",
        "{\"client_name\":\"CloudAEye JetBrains\",\"application_type\":\"native\",\"token_endpoint_auth_method\":\"none\",\"scope\":\"mcp:tools openid offline_access\"}"
      ]
    }
  }
}

Click OK > Apply, complete browser sign-in, and check the connection's Status and tool list. The inline metadata describes a public native client; it contains no credentials. On macOS/Linux, use npx instead of npx.cmd.

Do not keep duplicate HTTP and STDIO connections. See JetBrains MCP setup.

Tools > MCP Server > Enable MCP Server exposes IntelliJ IDEA's own tools. It does not connect to CloudAEye; use the AI Assistant settings above.

3. Run a skill

Open View > Tool Windows > AI Assistant (or AI Chat). Select Claude Agent or Codex, send a message, and complete agent activation. Agent login and CloudAEye OAuth are separate. See agent activation and AI Chat.

In AI Assistant > Agents, enable Pass custom MCP servers where exposed, then start a new chat. Pass IntelliJ MCP server refers to the IDE's separate server. See agent settings.

Type $ and select init, then describe. Use the picker name rather than assuming a cloudaeye: namespace; Try in chat also works.

Run describe on a meaningful uncommitted change and wait for the CloudAEye result. Then install review to analyze uncommitted changes or an eligible PR. See the shared workflow.

3.1 Help and Commands

For the complete command catalog, run CloudAEye's help or see Skills & Commands reference guide.

Troubleshooting

Problem Action
Skills page missing Update IntelliJ IDEA and AI Assistant.
Skills already in Installed Enable them and use Try in chat.
Repeated 401 Check bridge logs, OAuth completion, client metadata, and endpoint.
Agent cannot see MCP tools Enable Pass custom MCP servers where supported; start a new chat.
Helper or alias fails Check the compatibility context; do not bypass restrictions.
Agent cannot send messages Complete agent activation or restore its usage allowance.

For MCP logs, use Help > Show Log in Explorer and inspect the mcp folder. Remove secrets before sharing logs.