cachly in JetBrains IDEs

Name: cachly
Author: cachly

The plugin shows your Brain inside the IDE. Your AI assistant uses the Brain through MCP (Model Context Protocol, the way assistants call outside tools). The plugin writes that connection for you where the assistant reads it from a file.

What the plugin does

Gets you a Brain on first start
With no key stored, the plugin asks for a free trial Brain by itself. No account, no key to paste. It then makes one request with the new key and shows the result in the status bar.
Status bar
Shows how many lessons your Brain holds. It refreshes every 300 seconds; you can change that in Settings → Tools → Cachly Brain.
Briefing on project open
If your Brain already holds lessons, a notification shows the three most reused ones.
Warning on file open
When you open a file, the plugin asks your Brain whether it knows a failure pattern for that path. It only speaks up for medium or high risk.
Tools menu
Connect Brain, Show Brain Health, Show Lessons, Save Lesson to Brain (Ctrl+Shift+B), Set Up AI Files and Run Brain Doctor.

Connect your AI assistant

The plugin alone does not give an assistant memory. The assistant needs the cachly MCP server. It runs through npx, so Node.js must be installed.

Claude Code.mcp.json

Run Tools → Set Up AI Files. The plugin writes .mcp.json into the project root. Claude Code reads it when it starts in that folder.

Junie.junie/mcp/mcp.json

Run Tools → Set Up AI Files (plugin 0.8.3 or newer). The plugin writes .junie/mcp/mcp.json, the project-level MCP file Junie reads.

JetBrains: Junie MCP configuration

JetBrains AI AssistantIDE settings

AI Assistant reads MCP servers from its settings, not from a project file. Open Settings → Tools → AI Assistant → Model Context Protocol (MCP), add a server and paste the JSON below. Choose whether it applies globally or to this project only.

JetBrains: AI Assistant and MCP

The server entry

This is what Set Up AI Files writes. For AI Assistant, paste it by hand. Your API key and instance ID are in Settings → Tools → Cachly Brain.

{
  "mcpServers": {
    "cachly": {
      "command": "npx",
      "args": ["-y", "@cachly-dev/mcp-server@latest"],
      "env": {
        "CACHLY_API_URL": "https://api.cachly.dev",
        "CACHLY_JWT": "<your API key, cky_…>",
        "CACHLY_BRAIN_INSTANCE_ID": "<your instance ID>"
      }
    }
  }
}

Both files contain your API key. If the repository is shared, add .mcp.json and .junie/mcp/mcp.json to .gitignore.

If the status bar says "not configured" or "offline"

Run Tools → Run Brain Doctor. It checks the key, the instance ID and the connection, and names the step that fails. If no key is stored, Tools → Connect Brain starts the setup again.

More: Plugin on JetBrains Marketplace · IDE plugins overview · MCP integration