JetBrains MCP Setup: Connect AgentReady to IntelliJ, Fleet, and AI Assistant

Most MCP setup guides focus on VS Code, Cursor, and Claude Desktop. But JetBrains has over 14 million active users across IntelliJ IDEA, WebStorm, PyCharm, GoLand, and Fleet — and JetBrains AI Assistant now supports MCP servers. If you use any JetBrains IDE, you can connect AgentReady and query your documentation directly from within the assistant.

This guide covers the setup for JetBrains AI Assistant (available in all major JetBrains IDEs) and Fleet, JetBrains's next-generation code editor.

JetBrains AI Assistant and MCP

JetBrains AI Assistant is the built-in AI chat panel in IntelliJ IDEA 2025.1+ and all recent JetBrains IDEs. It supports two modes:

— Chat mode — a persistent AI assistant panel you can ask questions in, similar to Claude Desktop's chat interface

— Agent mode — an agentic flow where the AI can take actions in your project with tool calls

MCP servers are used in both modes. When AgentReady is connected, the AI can call ask_site during a conversation — for instance, when you ask "how does Stripe handle idempotency keys?" — and get a cited answer from the indexed docs rather than synthesizing from training data.

Setup: IntelliJ IDEA and other JetBrains IDEs

JetBrains AI Assistant uses a JSON configuration file for MCP servers, similar to Claude Desktop.

Step 1 — Find the config file. The location depends on your OS:

# macOS
~/Library/Application Support/JetBrains/AIAssistant/mcp.json

# Linux
~/.config/JetBrains/AIAssistant/mcp.json

# Windows
%APPDATA%JetBrainsAIAssistantmcp.json

Create the file if it doesn't exist yet.

Step 2 — Add AgentReady. There are two options: the remote HTTP endpoint (no local install) or the stdio bridge via npx.

{
  "mcpServers": {
    "agentready": {
      "command": "npx",
      "args": ["-y", "@agentreadyweb/mcp"],
      "env": {}
    }
  }
}

If you prefer the remote HTTP transport (no npx required):

{
  "mcpServers": {
    "agentready": {
      "url": "https://www.agentready.it.com/api/mcp",
      "transport": "streamable-http"
    }
  }
}

Step 3 — Restart the IDE. JetBrains AI Assistant reads the config on startup. After restarting, open the AI Assistant panel (View → Tool Windows → AI Assistant) and you should see AgentReady listed under available tools.

Step 4 — Test it. In the AI Assistant chat panel:

You: Using the ask_site tool, what does stripe.com/docs say about
     webhook signature verification?

AI Assistant: [calls ask_site] Based on docs.stripe.com/webhooks/signatures:
"Stripe signs webhook events using a signing secret unique to each
endpoint. Verify the Stripe-Signature header using
stripe.webhooks.constructEvent()..."
Source: docs.stripe.com/webhooks/signatures

Setup: Fleet

Fleet is JetBrains's newer, lighter editor. Its AI configuration lives in a workspace settings file rather than a global config.

Create or edit .fleet/settings.json in your workspace root:

{
  "ai.mcp.servers": {
    "agentready": {
      "command": "npx",
      "args": ["-y", "@agentreadyweb/mcp"]
    }
  }
}

Fleet picks up the MCP config when the workspace loads. Open the AI panel (⌘K on macOS, Ctrl+K on Linux/Windows) and AgentReady will be available as a tool.

Which docs to connect first

The highest-value sites to have available in your JetBrains AI Assistant depend on your stack. The AgentReady index already includes most common developer documentation:

— Backend devs: Spring Boot docs, Hibernate, Ktor, Quarkus, FastAPI, Django — all indexed

— Frontend devs: React, Vue, Next.js, Vite, TypeScript — all indexed

— DevOps/infra: Kubernetes, Docker, Terraform, AWS, CloudFlare — all indexed

— Your own docs: Index your internal or product docs at agentready.it.com — they'll be queryable alongside the public index

Use list_sites to see the full index, or submit a new domain with submit_site if what you need isn't there yet.

Comparing JetBrains to VS Code and Cursor

The MCP experience in JetBrains AI Assistant is functionally equivalent to Cursor's — both use the same protocol, same tools, same endpoint. The practical differences are in the IDE workflows around it.

In IntelliJ specifically, the AI Assistant is more tightly integrated with Java/Kotlin refactoring than VS Code-family editors. When you ask about a library, the context includes your open files, so AgentReady's cited answer can be combined with class-level awareness of your codebase. That combination — "what does the docs say about X" + "here's how X appears in your code" — is where JetBrains AI tends to produce useful suggestions.

Troubleshooting

AI Assistant doesn't show AgentReady tools. Check that the mcp.json is valid JSON (no trailing commas), that Node.js is in your PATH (for the npx variant), and that you've fully restarted the IDE — not just reloaded the project.

npx takes too long on first run. The first invocation downloads the package. Switch to the remote HTTP transport for a zero-install setup.

Fleet doesn't see the config. Ensure .fleet/settings.json is in the workspace root (the folder you opened in Fleet, not a subfolder). Fleet doesn't search parent directories for settings.