JetBrains configures MCP servers through the AI Assistant settings UI instead of a repo file. Paste the native stdio JSON block below into the MCP dialog.
OVERVIEW
JetBrains is UI-driven. The MCP server is created from Settings rather than from a dotfile in the repo, but for Zephex the reliable shape is the stdio block below running `npx -y zephex` with `ZEPHEX_API_KEY` in the env block.
CONFIG FILE LOCATION
Open Settings -> Tools -> AI Assistant -> Model Context Protocol (MCP), then add a new server and paste the JSON below.
Use the JSON exactly as shown and keep the server name as zephex.
Replace `mcp_sk_your_key_here` with the real key from Dashboard -> API Keys.
{ "mcpServers": { "zephex": { "command": "npx", "args": ["-y", "zephex"], "env": { "ZEPHEX_API_KEY": "mcp_sk_your_key_here" } } }}JetBrains should show the hosted MCP server in the AI Assistant MCP settings before you rely on it in normal chats.
The IDE usually was not restarted after the server was created. Restart the IDE fully, then reopen the AI Assistant panel.
Check that the JSON uses `command`, `args`, and `env` exactly as shown above instead of the older HTTP URL/header fields.
If you are using the JSON as a reference, make sure the JetBrains UI fields map to the same values instead of pasting malformed JSON into a free-text field.
Confirm the AI Assistant plugin is enabled and that you added the server under MCP Servers in the active IDE, not in a different JetBrains product on the same machine.
If the editor still does not connect, return to Quickstart or check Connection Issues.