Originally published at devtoolpicks.com
Cursor launched version 3 on April 2, 2026, and it is a bigger shift than most point releases. This is not an incremental update. The team rebuilt the interface from scratch around a new idea: that most code will be written by AI agents, and the developer's job is to orchestrate them, not write every line.
If you are currently paying for Cursor, or deciding between Cursor, Claude Code, and GitHub Copilot, this changes the conversation.
Here is what actually shipped, what it costs, and whether the verdict changes for solo developers and indie hackers.
What Is Cursor 3?
Cursor 3 is the third major version of the AI coding tool built by Anysphere. The company has raised over $3 billion from Nvidia, Google, and others, and hit a reported $2B ARR in early 2026. The new version was built under the internal codename "Glass" and is Cursor's direct answer to Anthropic's Claude Code and OpenAI's Codex, both of which have been eating into Cursor's share of the agentic coding market.
The official framing from the Cursor team: "We are introducing Cursor 3. It is simpler, more powerful, and built for a world where all code is written by agents, while keeping the depth of a development environment."
That is not marketing copy. The interface genuinely reflects this shift.
What Actually Changed in Cursor 3
The Agents Window
The biggest change. Cursor 3 replaces the Composer pane with a dedicated Agents Window, a full-screen workspace designed for running and managing multiple AI agents simultaneously.
Previously you had one chat, one agent, one task at a time. Now you run as many parallel agents as you want, each handling different tasks across different repositories or environments. Each agent gets its own Agent Tab, and you can view them side-by-side or in a grid layout. Every agent, whether local or running in the cloud, appears in a unified sidebar.
To try it: update Cursor and hit Cmd+Shift+P then type "Agents Window". You can switch back to the IDE at any point, or run both simultaneously.
Design Mode
New in Cursor 3. When you are working on a UI, you can activate Design Mode to annotate browser elements directly. Instead of describing in text which part of the interface you want changed, you click and select the element, then add your instruction. The agent implements the change from there.
Toggle it with Cmd+Shift+D. Use Shift+drag to select an area, Cmd+L to add an element to the chat. For frontend work, this closes the gap between Cursor and dedicated tools like v0.
Cloud and Local Agent Handoff
Agents now move between cloud environments and your local machine without friction. You can start a task on cloud infrastructure, then hand it off to a local agent for finishing and review. Cloud agents automatically produce screenshots and demo videos of their work so you can verify what they did without running it yourself.
Cloud agents can also be launched from outside the IDE: Slack, GitHub, Linear, mobile devices, or the web. An agent can be kicked off from a GitHub issue and you can check its progress from your phone.
Built-In Git
Git is now native to the Cursor interface. Staging, committing, and creating pull requests happen directly inside Cursor 3 without switching to a terminal or external tool. The entire flow from code change to merged PR can stay inside one window.
/worktree and /best-of-n Commands
Two new commands worth knowing:
/worktree runs a task in an isolated Git worktree, keeping experimental work separate from your main branch automatically.
/best-of-n runs the same prompt against multiple AI models and presents the results side by side so you can pick the best output. Useful when you are not sure which model handles a specific problem better.
Pricing in 2026
Cursor has six tiers as of April 2026:
| Plan | Cost | What You Get |
|---|---|---|
| Hobby | Free | Limited Tab completions, limited agent requests |
| Pro | $20/month ($16 annual) | Unlimited Tab, $20 credit pool, cloud agents |
| Pro+ | $60/month | 3x credits ($60 pool), same features as Pro |
| Ultra | $200/month | 20x credits, priority feature access |
| Teams | $40/user/month | Pro features + admin controls, SSO, centralised billing |
| Enterprise | Custom | Pooled usage, dedicated support, compliance |
The credit system matters. Every paid plan includes a monthly credit pool equal to the plan price. Auto mode is unlimited and does not draw from this pool. Manually selecting a premium model (Claude Sonnet, GPT-4o) draws from your credits based on token consumption. A $20 Pro pool covers roughly 200-250 Sonnet requests, or significantly more if you stay on Auto.
For solo developers, Pro at $20/month is the right tier for most. Use Auto mode as the default and save manual model selection for complex, multi-file problems.
The credit system caught some developers off guard when it launched in mid-2025 because it replaced the simpler 500-fast-requests model. The practical impact: if you run Cursor in Auto mode for normal daily coding, you will rarely exhaust the pool. If you manually select frontier models for every chat message, you will.
Does Cursor 3 Change the Verdict vs Claude Code and Copilot?
The short answer: Cursor is now more competitive with Claude Code on agentic work, but the choice still comes down to your workflow.
If you want the full breakdown with pricing comparisons across all three tools, the Cursor vs GitHub Copilot vs Claude Code comparison covers it in detail. Here is how Cursor 3 specifically shifts the picture:
Cursor 3 vs Claude Code: Claude Code has been ahead on fully autonomous, terminal-based agentic work. The Agents Window closes this gap significantly. Parallel agents, cloud execution, and mobile launch support put Cursor on similar ground for orchestrating complex tasks. Claude Code still wins for developers who prefer a terminal-first workflow and do not want a full IDE. Cursor 3 wins for developers who want agent power inside a familiar editor with a GUI.
Cursor 3 vs GitHub Copilot: Copilot has been cheaper ($10/month vs $20/month for Pro) but less capable on agentic features. Copilot introduced its own coding agent in early 2026, but it is less mature than Cursor's. The gap has widened with Cursor 3. If you are on Copilot and hitting its limits on multi-file or agentic tasks, Cursor 3 gives you a meaningful upgrade. If Copilot's inline suggestions are all you actually use day-to-day, the $10 price difference is hard to justify.
Who should stay on Cursor: Anyone who already uses Composer or Agent mode regularly. The Agents Window is a direct upgrade to that workflow and the migration is automatic on update.
Who might reconsider: Solo developers who only use Cursor for Tab completions and light chat. At that usage level, $20/month for Cursor versus $10/month for Copilot is a real difference with no functional gap.
Should You Upgrade?
If you are already on Cursor Pro, the update is automatic and free. There is no price increase for Cursor 3. Open the app, update, and the Agents Window is available on demand.
If you are on the free Hobby plan evaluating whether to pay, Cursor 3 makes the Pro tier more defensible. Parallel agents, cloud execution, Design Mode, and built-in Git are all behind the $20/month paywall. The Hobby plan still exists and is still a genuine free option for evaluation, not a nag screen.
If you are currently on Claude Code or Copilot and wondering whether to switch: Cursor 3 brings more feature depth to the IDE experience than either competitor offers right now. The tradeoff is price. Claude Code at $20/month on Claude Pro is comparable cost to Cursor Pro but feels more terminal-native. Copilot at $10/month is cheaper but behind on agents.
FAQ
Is Cursor 3 a free update?
Yes, for existing subscribers. If you are on any paid Cursor plan, update the app and Cursor 3 is there. Activate the Agents Window via Cmd+Shift+P then "Agents Window". Pricing has not changed with this release.
What happened to Composer?
Composer is replaced by the Agents Window. The Agents Window is the evolution of Composer: same idea, rebuilt for parallel agents and cloud execution. You can still use the traditional IDE view and switch between them.
Can I still use Cursor as a normal IDE?
Yes. The classic IDE layout is still available. Cursor 3 adds the Agents Window as an option, not a replacement. You can have the IDE and the Agents Window open simultaneously, or ignore the Agents Window entirely if you only use Tab completions and Chat.
Does Cursor 3 support Laravel and PHP projects?
Yes. Cursor works with any language or framework. The agentic features, including multi-repo support and worktrees, work with PHP and Laravel projects the same as any other codebase.
What is the /best-of-n command?
A new command in Cursor 3 that runs the same prompt against multiple AI models and returns the outputs side by side. Useful when you want to compare how Claude Sonnet, GPT-4o, and Cursor's own model handle the same problem before deciding which output to accept.
The Bottom Line
Cursor 3 is a real release. The Agents Window, parallel execution, cloud agent handoff, Design Mode, and built-in Git are all meaningful additions that move Cursor from a smart IDE with AI toward something closer to an agent orchestration platform that happens to have an IDE attached.
For solo developers and indie hackers who build with AI tools daily, this is the most capable version of Cursor yet. At $20/month for Pro, the price has not changed. The capabilities have.
The full three-way comparison with pricing and use case breakdowns: Cursor vs GitHub Copilot vs Claude Code in 2026.
Top comments (0)