Hey Prompt Warrior,

Cursor just dropped some huge updates that are about to completely change how I code.

These features are pushing us from the co-pilot era into true AI agent territory—it's becoming more and more like having a team of developers working for you 24/7.

In this post, I'll walk you through each new feature and show you exactly how to use them to 10x your AI vibe coding game.

Here's what we'll cover:

• Background agents that code while you sleep

• BugBot: Your AI code reviewer

• Simplified MCP tool integrations

• Memory features 

• and more

Let's dive in!

Read time: 5 minutes

🤖 Background Agents: Your Autonomous Coding Team

This is the update I'm most excited about, and here's why it matters.

Until now, Cursor's agent feature worked in co-pilot mode. You'd type a prompt, watch the AI write code, and approve each change. Great, but you're still babysitting the process.

Background agents flip this completely. Now I can delegate entire features and walk away. (I’m just doing small and easy features for now, but as the models improve I can up the difficulty).

How It Works:

When you open Cursor, you'll see a new "Background" option in the dropdown. Here's what I did with it:

The Task: Create a pricing landing page (something that's been sitting in my backlog forever).

My Prompt:

I basically wrote a small specification—the kind of ticket I'd normally hand to a junior developer.

What Happened Next:

• The background agent went off and built the entire feature

• Took about 5 minutes for this simple task

• Created a new branch (no messing with main!)

• Generated a complete pull request with all changes

When it finished, I got a notification with:

• A summary of what was created

• Option to review changes locally

• Direct link to the pull request

I reviewed the PR, tested it locally, and merged it straight to production. Done.

The Game-Changer Part:

You can run multiple background agents in parallel. While I'm working on complex features, I can spin up 3-4 agents handling smaller tasks. It's like having my own development team that never needs breaks.

Key Takeaway: This isn't just faster coding—it's a fundamental shift in how we work. I can now hand off my tedious tasks that were piling up in the backlog to background agents and focus on the challenging stuff.

🐛 BugBot: Your AI Code Reviewer

BugBot is another coding agent that lives inside of your GitHub, and you can call to review your code.

Setting Up BugBot:

1. Connect GitHub: Cursor makes this simple—just log in through their interface

2. Activate for Your Repo: I enabled it for my CopyCoder repository

3. Configure Settings: I set it to "only run when mentioned" to avoid spam

How to Use It:

On any pull request, just comment:

BugBot run

That's it. BugBot reviews the entire PR and either:

• Confirms no bugs found (like in my pricing page example)

• Lists specific bugs with explanations

It's like having a senior developer review every line of code, except it takes 30 seconds instead of 3 days.

Pro Tip: Use this on every PR before merging. It catches issues humans miss and gives you confidence in your deployments. BUT: Quite pricey, since it uses the Max mode.

🔧 MCP Tools: Integrations Made Simple

MCP (Model Context Protocol) tools are hot right now, and Cursor just made them much easier to use.

What Changed:

Before: It was a pain setting them up. You had to edit JSON files in settings, deal with configuration hell, pray it works.

Now: Click "Add" on any tool from their curated list. Done.

Real Example: Notion Integration

I added the Notion MCP in literally 30 seconds:

1. Found Notion in the MCP tools list

2. Clicked "Add Notion"

3. Hit "Install"

Now I can create tickets directly from Cursor:

• Assign to team members

• Add details and context

• Update ticket status

• All without leaving my editor

This is huge for maintaining flow state. No more context switching between tools.

Current Limitations: The list is small since they just launched, but expect this to explode with integrations for everything from Slack to Linear.

🧠 Memory Features: Your AI Actually Remembers

Cursor now added a memory feature similar to the memory feature that ChatGPT has.

Example:

Every time I asked Cursor to install Shadcn components, it used an outdated command. Every. Single. Time.

Now I just told it: "Remember that when you install Shadcn components, do not use the old outdated command, instead use the new command [specific command here]"

Cursor added this to its memory, and now it never makes that mistake again.

How to Use Memories:

• Tell Cursor to "remember" anything in chat

• Check your memories under the "Rules" tab

• Edit or delete memories as needed

📊 Enhanced Chat: Actually Useful Visualizations

The chat interface got a massive upgrade that makes understanding complex codebases so much easier.

What's New:

Mermaid Diagrams: Ask for a visualization of your architecture and get an actual diagram in chat.

Rich Tables: No more ugly text tables. Get properly formatted data displays.

My Favorite Use Case:

"Draw a mermaid diagram of how our APIs interact"

Or: "List all our APIs in a table with their purposes"

Clean, scannable output that beats scrolling through files.

🎯 TLDR

We're moving from AI as a co-pilot to AI as an autonomous team member. Background agents handle the grunt work, BugBot catches mistakes, and integrated tools keep you in flow.

The productivity gains are real. Tasks that took hours now happen in the background while I focus on architecture and user experience.

My Prediction: In 6 months, coding without these features will feel like coding without autocomplete—technically possible, but why would you?

Thanks for reading!

If you enjoyed this, consider forwarding this newsletter to a friend or colleague.

What would you like me to write about in the next post? Reply to this email and let me know.

Moritz

P.S. Are you a beginner and want to learn how to start building software applications with Cursor?

That’s exactly what we do in the Prompt Warrior community.

You’ll find in-depth guides, step-by-step courses, ready-to-use templates, and live Q&A sessions where you can get direct feedback.