Video tutorial · The Metaverse Guy
Cursor 2.0 Review: Multi-Agent Test & Honest Verdict
I test Cursor 2.0's multi-agent workflows on a React AI thumbnail generator and a Python RAG agent, comparing it to Windsurf and Copilot.
- Published
- 30 October 2025
- Length
- 9 min
- Views
- 22,214 views
Cursor 2.0 brings multi-agent workflows, a new Composer model, and an embedded browser. I put it through its paces by building a React project with Google AI Studio and testing a Python RAG agent, where I hit a configuration error. Here's what worked, what didn't, and how it stacks up against Windsurf and GitHub Copilot.
First Impressions & Multi-Agent Setup
Cursor 2.0 has been hyped as a game-changer, but I wanted to see if it actually helps ship real apps faster. The standout feature is multi-agent workflows: you can run up to 8 agents in parallel, each tackling a different part of your project. The interface lets you spin them up quickly, and in theory they can work on imports, dependency installs, and code generation simultaneously.
Python RAG Agent: The 'No LLM configured' Error
I first tried to build a Python RAG agent inside Cursor. Almost immediately, I hit a wall: the agent kept throwing a 'No LLM configured' error. It wasn't clear how to resolve it within the multi-agent flow, and the context seemed to get lost. This highlighted that for more complex, non-trivial setups, Cursor 2.0 can struggle with configuration and context retention.
Building a React AI Thumbnail Generator
I switched to a React project—an AI thumbnail generator using Google AI Studio (Gemini) and Nanobanana. This time, the process was much smoother. I imported the code, installed dependencies, and ran the dev server without major issues. The key step was creating a .env.local file and adding my Gemini API key so the app could call the model.
- Import the existing React project code into Cursor.
- Install dependencies (npm install or yarn).
- Run the dev server with npm run dev and check the local preview.
- Create a .env.local file and add the Gemini API key for Nanobanana.
- Test the thumbnail generation—it worked as expected.
Parallel Agents: Code Review & Bug Scan
While the React app was running, I spun up a second agent to do a code review and bug scan. The agent analysed the codebase and flagged potential issues. Having two agents work side by side—one building, one reviewing—felt productive, though I noticed the review agent's suggestions were sometimes generic. It's a promising workflow if you keep tasks clearly scoped.
What's New & How It Compares
Beyond multi-agent, Cursor 2.0 introduces a new Composer coding model, an embedded browser for previewing apps, and a refined code diff UX. These make the editor feel more integrated. When comparing to Windsurf and GitHub Copilot, Cursor's parallel agents are a differentiator, but I found Windsurf more stable for longer sessions and Copilot better at understanding large codebases. Cursor's speed is good for quick React setups, but context handling on complex Python tasks still needs work.
Verdict: Who Should Use Cursor 2.0
Cursor 2.0 is worth trying if you frequently scaffold React projects or want to experiment with parallel AI agents. It's not yet a replacement for more stable tools on complex, multi-file Python or RAG applications. My recommendation: use Cursor for rapid prototyping and pair it with another assistant like Copilot or Windsurf for deeper debugging. Keep an eye on future updates—the multi-agent idea has real potential once the rough edges are smoothed out.
Tools used in this video
- Cursor 2.0
- Google AI Studio
- Gemini
- Nanobanana
- React
- Python
- Windsurf
- GitHub Copilot
- Claude
Questions
- What is Cursor 2.0's multi-agent feature?
- It lets you run up to 8 AI agents in parallel, each working on different tasks like installing dependencies, writing code, or reviewing changes simultaneously.
- How does Cursor 2.0 compare to Windsurf and GitHub Copilot?
- Cursor offers unique parallel agents, but Windsurf tends to be more stable for long sessions, and Copilot often handles large codebases with better context. Cursor is fastest for quick React setups.
- Can I use Cursor 2.0 for Python RAG projects?
- In my test, a Python RAG agent hit a 'No LLM configured' error and wouldn't run. It may require additional configuration, so it's not seamless out of the box for complex Python tasks.
- Is Cursor 2.0 free?
- The video doesn't cover pricing, but the description mentions looking for a free Cursor alternative, suggesting the full version may not be free. Check the official site for current plans.
Chapters
- 00:00 Hot take: Cursor 2.0 hype vs reality
- 00:48 Multi-Agent interface (parallel agents explained)
- 01:35 Python RAG agent test → “No LLM configured” error
- 02:18 Switching to React project (AI Thumbnail Generator)
- 03:02 Install & run: deps, npm run dev, local preview
- 03:44 .env.local + Gemini API key (Nanobanana)
- 04:20 Does it actually work? Generating the thumbnail
- 05:05 Spinning a second agent: code review / bug scan
- 05:46 What’s new: Composer model, embedded browser, code diff UX
- 06:40 Cursor vs Windsurf vs Copilot (context, stability, speed)
- 07:40 Pros & cons: who should use Cursor 2.0 right now
- 08:25 Final verdict & tool stack recommendation
Video description and links
This article was written from the video's description with the help of AI. Watch the video for the full walkthrough.



