Training Example: Cursor – Review the Data, Give Your Score & Compare to the Real AI Evaluation

[H1] Cursor is your coding agent for building ambitious software.The agent-native way to build ambitious software.
Download for macOS⤓Get started→Request a demo→This element contains an interactive demo for sighted users showing multiple Cursor interfaces: Cursor Desktop interface, Cursor CLI interface. The interface is displayed over a subtle, solid brand background.Cursor DesktopGet Cursorhttp://localhost:3000Software creation is changing. We are a group of researchers, engineers, and technologists inventing at the edge of what's useful and possible.We have much to learn, try, and build.See projectsJoin our team →2026Secure codebase indexingPublished2026Semantic searchPublished2025Reinforcement learningPublished2024Shadow workspacesPublished2024Multi-agent collaborationPublished2023Context-aware completionsPublished2023Intelligent code navigationPublished2022Fast inference optimizationPublished2022Language model fine-tuningPublishedBuild Landing Page make a landing page based on attached docs explaining what we doCursor CLIGet CLICursor Agent~/cursor/cursor-webAnalyze Tab vs Agent Usage PatternsHelp me understand how teams split their focus between the tab view and the agents panel across our workspaces.→Composer 2.5/ for commands · @ for files
[H2] Trusted every day by teams that build world-class software
[H3] Agents turn ideas into code
Accelerate development by handing off tasks to Cursor, while you focus on making decisions.Learn about agentic development →This element contains an interactive demo for sighted users. It's a demonstration of Cursor's IDE showing AI-powered coding assistance features. The interface is displayed over a subtle, solid brand background.CursorGet Cursorfeature-prd.mdpresence.tsPlansfeature-prd.mdComposer 2.5Build
[H2] Mission Control Interface
A grid view of all open windows as scaled previews, allowing quick selection to bring any window to front.
[H2] Trigger
Menu item in MenuBar.tsx (View > Mission Control), hotkey F3, or double-tap desktop.
[H2] View Behavior
Overlay existing windows into a grid of live previews with spring-based layout animations and shared element transitions.3 TasksAdd multiplayer mode to useAppStore.tsCreate a new MissionControlView.tsx componentUpdate AppManager.tsx to apply expose modes.Add a task, ⌘K to generate...Plan Mission Control let's build a mission control interface, similar to the expose-style window manager on macOS
[H3] Works autonomously, runs in parallel
Agents use their own computers to build, test, and demo features end to end for you to review.Learn about cloud agents →cursor.com/agentThis WeekAcme Research DashboardLive Telemetry PipelineZero-Downtime DeploysThis MonthBinary Protocol ParserEdge Cache InvalidationAuth Token RotationAcme Research Dashboardlet's build a dashboard to make our research findings interactiveExplored 12 files, 4 searchesOn it. I'll build the dashboard using your theme config, wire up the research data, and add interactive charts with public access controls.Worked for 14m 22sProcessed screen recordingDone! Here's a walkthrough of the dashboard.Acme LabsSummaryBuilt the interactive dashboard with realtime charts, data from Snowflake, and shadcn components. Deployed to staging via Vercel.Add a follow up...AgentOpus 4.8
[H3] In every tool, at every step
Cursor runs in your terminal, collaborates in Slack, and reviews PRs in GitHub.curl https://cursor.com/install -fsS | bashThis element contains an interactive demo for sighted users showing multiple Cursor interfaces: the CLI with command-line assistance, Slack integration for team communication. The interface is displayed over a scenic painted landscape wallpaper, giving the demo an artistic backdrop.SlackGet Cursor for Slack
[H2] #feature-realtime-sync
8 members
[IMG: member-1]
[IMG: member-2]
[IMG: member-3]
[IMG: member-4]
[IMG: swhitmore avatar]
swhitmore5mi wanna be able to go to cursor.com/changelog#1.0 to see 1.0 changelog
[IMG: eric avatar]
eric4mchecks out@cursor can you take a stab?
[IMG: Cursor avatar]
CursorAPP2mI added direct linking for changelog entries and updated the Node.js version constraints across the project for better maintainability.View PROpen in CursorOpen in Webcursor-agentGet CLI> agentCursor Agent~/anysphere/research · main→Plan, search, build anythingGPT-5.5 Extra High Fast/ commands · @ files · ! shell
[H3] Magically accurate autocomplete
Our specialized Tab model predicts your next action with striking speed and precision.Learn about Tab →
[IMG: Abstract waves background]
This element contains an interactive demo for sighted users. It's a demonstration of Cursor's IDE showing AI-powered coding assistance features. The interface is displayed over a scenic painted landscape wallpaper, giving the demo an artistic backdrop.CursorGet CursorDashboard.tsxSupportChat.tsx"use client";
import React, { useState } from "react";
import Navigation from "./Navigation";
import SupportChat from "./SupportChat";
export default function Dashboard() {
return (
<div className="flex h-[600px] border rounded-lg overflow-hidden">
<div className="w-64 border-r">
</div>
<div className="w-80 border-l">
<SupportChat />
</div>
</div>
);
}"use client";
import React, { useState } from "react";
import Navigation from "./Navigation";
import SupportChat from "./SupportChat";
export default function Dashboard() {
return (
<div className="flex h-[600px] border rounded-lg overflow-hidden">
<div className="w-64 border-r">
</div>
<div className="w-80 border-l">
<SupportChat />
</div>
</div>
);
}
[H2] The new way to build software.
“It was night and day from one batch to another, adoption went from single digits to over 80%. It just spread like wildfire, all the best builders were using Cursor.”Diana Hu General Partner, Y Combinator“My favorite enterprise AI service is Cursor. Every one of our engineers, some 40,000, are now assisted by AI and our productivity has gone up incredibly.”Jensen Huang President & CEO, NVIDIA“The best LLM applications have an autonomy slider: you control how much independence to give the AI. In Cursor, you can do Tab completion, Cmd+K for targeted edits, or you can let it rip with the full autonomy agentic version.”Andrej Karpathy CEO, Eureka Labs“Cursor quickly grew from hundreds to thousands of extremely enthusiastic Stripe employees. We spend more on R&D and software creation than any other undertaking, and there's significant economic outcomes when making that process more efficient.”Patrick Collison Co‑Founder & CEO, Stripe“The most useful AI tool that I currently pay for, hands down, is Cursor. It's fast, autocompletes when and where you need it to, handles brackets properly, sensible keyboard shortcuts, bring-your-own-model... everything is well put together.”shadcn Creator of shadcn/ui“It's definitely becoming more fun to be a programmer. We are at the 1% of what's possible, and it's in interactive experiences like Cursor where models like GPT-5 shine brightest.”Greg Brockman President, OpenAI
[H2] Stay on the frontier
[H3] Use the best model for every task
Choose between every cutting-edge model from OpenAI, Anthropic, Gemini, xAI, and Cursor.Explore models ↗Ask Cursor to plan or build anythingAgentAutoAutoSuggested✓Composer 2.5GPT-5.5Opus 4.8Gemini 3.1 ProGrok 4.3
[H3] Complete codebase understanding
Cursor learns how your codebase works, no matter the scale or complexity.Learn about codebase indexing ↗Where are these menu label colors defined?
[H3] Develop enduring software
Trusted by over half of the Fortune 500 to accelerate development, securely and at scale.Explore enterprise →
[H2] Changelog
3.8Jun 18, 2026Improvements to Cursor Automations3.7Jun 17, 2026Cloud Environment Setup and Cloud Subagents in Agents WindowJun 10, 2026Bugbot is now over 3x faster, 22% cheaper, and finds 10% more bugs3.7Jun 5, 2026Design Mode ImprovementsSee what's new in Cursor →
[H3] Cursor is an applied research team focused on building the future of software development.
Join us→
[H2] Recent highlights
Mar 27, 2026·ResearchA technical report on Composer 2Sasha Rush · 3 min readMay 18, 2026·ResearchIntroducing Composer 2.5Cursor Team · 7 min readApr 2, 2026·ProductMeet the new CursorMichael & Sualeh · 10 min readJun 5, 2026·ProductDirect agents with visual prompts in Design ModeErik, Ian & Ryo · 6 min readView all blog posts →
[H2] Try Cursor now.
Download for macOS⤓Get started→

[IMG: Abstract waves background]
Acme Research Dashboardbuild a dashboard to visualize our findingsExplored 12 files, 4 searchesOn it. I'll build the dashboard using your theme config, wire up the research data, and add interactive charts with public access controls.Worked for 14m 22sProcessed screen recordingDone! Here's a walkthrough of the dashboard.Acme LabsSummaryBuilt the interactive dashboard with realtime charts, data from Snowflake, and shadcn components. Deployed to staging via Vercel.Add a follow up...AgentComposer 2.5ResearchDashboard.tsxresearchDataset.tsimport { useMemo } from "react";import { InsightCard } from "./InsightCard";import { FindingsChart } from "./FindingsChart";import { ResearchFilters } from "./ResearchFilters"; type Finding = { id: string; label: string; confidence: number; trend: number[]; category: "product" | "market" | "ops";}; export function ResearchDashboard({ findings,}: { findings: Finding[];}) { const summary = useMemo(() => { return findings.reduce( (acc, finding) => { acc.total += 1; acc.avgConfidence += finding.confidence; if (finding.category === "product") acc.product += 1; return acc; }, { total: 0, avgConfidence: 0, product: 0 }, ); }, [findings]);  const avgConfidence = summary.total > 0 ? Math.round(summary.avgConfidence / summary.total) : 0;  return ( <section className="grid gap-6"> <header className="flex items-end justify-between"> <div> <h1 className="text-2xl font-semibold">Research Dashboard</h1> <p className="text-sm text-slate-500"> Explore findings, confidence, and rollout readiness. </p> </div> <ResearchFilters /> </header>  <div className="grid grid-cols-3 gap-4"> <InsightCard label="Findings" value={summary.total} /> <InsightCard label="Avg confidence" value={`${avgConfidence}%`} /> <InsightCard label="Product signals" value={summary.product} /> </div>  <div className="rounded-2xl border border-slate-200 p-5"> <FindingsChart series={findings.map((finding) => ({ id: finding.id, label: finding.label, points: finding.trend, }))} /> </div> </section> );}Acme Research Dashboardbuild a dashboard to visualize our findingsExplored 12 files, 4 searchesOn it. I'll build the dashboard using your theme config, wire up the research data, and add interactive charts with public access controls.Worked for 14m 22sProcessed screen recordingDone! Here's a walkthrough of the dashboard.Acme LabsSummaryBuilt the interactive dashboard with realtime charts, data from Snowflake, and shadcn components. Deployed to staging via Vercel.Add a follow up...AgentComposer 2.5ResearchDashboard.tsxresearchDataset.tsimport { useMemo } from "react";import { InsightCard } from "./InsightCard";import { FindingsChart } from "./FindingsChart";import { ResearchFilters } from "./ResearchFilters"; type Finding = { id: string; label: string; confidence: number; trend: number[]; category: "product" | "market" | "ops";}; export function ResearchDashboard({ findings,}: { findings: Finding[];}) { const summary = useMemo(() => { return findings.reduce( (acc, finding) => { acc.total += 1; acc.avgConfidence += finding.confidence; if (finding.category === "product") acc.product += 1; return acc; }, { total: 0, avgConfidence: 0, product: 0 }, ); }, [findings]);  const avgConfidence = summary.total > 0 ? Math.round(summary.avgConfidence / summary.total) : 0;  return ( <section className="grid gap-6"> <header className="flex items-end justify-between"> <div> <h1 className="text-2xl font-semibold">Research Dashboard</h1> <p className="text-sm text-slate-500"> Explore findings, confidence, and rollout readiness. </p> </div> <ResearchFilters /> </header>  <div className="grid grid-cols-3 gap-4"> <InsightCard label="Findings" value={summary.total} /> <InsightCard label="Avg confidence" value={`${avgConfidence}%`} /> <InsightCard label="Product signals" value={summary.product} /> </div>  <div className="rounded-2xl border border-slate-200 p-5"> <FindingsChart series={findings.map((finding) => ({ id: finding.id, label: finding.label, points: finding.trend, }))} /> </div> </section> );}
[H2] Desktop
An agent-first experience, with all the best parts of the editor for when you need them.Download for macOS⤓
[IMG: Sunlit woodland background]
Interactive demo showing Cursor Agent. The interface is displayed over a subtle, solid brand background.Cursor Agent> agentCursor Agent~/anysphere/research · main→Plan, search, build anythingGPT-5.5 Extra High Fast/ commands · @ files · ! shell
[H2] Terminal
Run agents in any terminal or script, with integrations to your preferred IDEs.curl https://cursor.com/install -fsS | bashAgentDashboardTodayVercel streaming SDK functionalityAnalyzing codebase·dashboard-uiModel selector dropdownPlanning next moves·stripe-landingImplement agent window2 models·agent-uiFix send button functionality2 Files+26-12·agent-uiThis WeekFix send button functionality2 Files+26-12·agent-uiImplement user authentication3 Files+45-8·auth-systemDesign responsive layout5 Files+30-3·frontendSet up database schema4 Files+35-2·databaseCreate API endpoints6 Files+50-5·backendAsk Cursor to build, plan, fix...
[H2] Web
Kick off agents in the cloud from your browser, phone, or tools like Slack, GitHub, and Linear.Start Cursor in Web→
[H3] 3.8Latest↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.7↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.6↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.5↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.4↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.3↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.2↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.1↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →
[H3] 3.0↓↑
[H3] macOS
Mac (ARM64)⤓Mac (x64)⤓Mac Universal⤓
[H3] Windows
Windows (x64) (System)⤓Windows (x64) (User)⤓Windows (ARM64) (System)⤓Windows (ARM64) (User)⤓
[H3] Linux
Linux .deb (ARM64)⤓Linux .deb (x64)⤓Linux RPM (ARM64)⤓Linux RPM (x64)⤓Linux AppImage (ARM64)⤓Linux AppImage (x64)⤓View release notes →

[H1] Turn ideas into code
Delegate implementation to focus on higher-level direction.Download for macOS⤓Try mobile agent→This element contains an interactive demo for sighted users showing multiple Cursor interfaces: Cursor Desktop interface, Cursor CLI interface. The interface is displayed over a subtle, solid brand background.Cursor DesktopGet Cursorfeature-prd.mdpresence.tsPlansfeature-prd.mdComposer 2.5Build
[H2] Mission Control Interface
A grid view of all open windows as scaled previews, allowing quick selection to bring any window to front.
[H2] Trigger
Menu item in MenuBar.tsx (View > Mission Control), hotkey F3, or double-tap desktop.
[H2] View Behavior
Overlay existing windows into a grid of live previews with spring-based layout animations and shared element transitions.3 TasksAdd multiplayer mode to useAppStore.tsCreate a new MissionControlView.tsx componentUpdate AppManager.tsx to apply expose modes.Add a task, ⌘K to generate...Plan Mission Control let's build a mission control interface, similar to the expose-style window manager on macOSCursor CLIGet CLICursor Agent~/cursor/cursor-webAnalyze Tab vs Agent Usage PatternsHelp me understand how teams split their focus between the tab view and the agents panel across our workspaces.→Composer 2.5/ for commands · @ for files
[H2] Everywhere you work
One agent across every surface.
[H2] Desktop
Manual to agentic coding, in one familiar editor.Download →Interactive demo with multiple windows showing Cursor's AI-powered features. The interface is displayed over a subtle, solid brand background.let's add a spotlight-style searchRead App.tsx, Menu.tsxSearched ⌘K shortcut handlersEdited Search.tsxDone. Press ⌘K to search files, actions, and settings.Ask Cursor to plan or build anythingAgentComposer 2.5
[H2] CLI
Run agents in any terminal, script, or editor.Install →Interactive demo with multiple windows showing Cursor's AI-powered features. The interface is displayed over a subtle, solid brand background.Cursor Agent~/cursor/cursor-webAnalyze Tab vs Agent Usage PatternsHelp me understand how teams split their focus between the tab view and the agents panel across our workspaces.⬢Thought 7s⬢Reviewed workspace usage exports and historical engagement notesI'll build an analytics toolkit that highlights tab versus agent adoption and switching behavior:summary.py+150-0Next I'll normalize the interaction labels and add session tagging so product can filter specific cohorts:segmentation.py+94-0Time to stitch everything together with a report helper that combines the key metrics:report.py+40-0I'll add tests to lock in the behaviour using a representative sample from last week's export:test_usage.py+90All set! We now track focus share, switching rates, and rolling engagement so PMs can compare tab-first and agent-first workflows in seconds.→Composer 2.5/ for commands · @ for files
[H2] Other Surfaces
Start agents from GitHub, Slack, Linear, JetBrains IDEs, and more.Add to Slack ↗Interactive demo with multiple windows showing Cursor's AI-powered features. The interface is displayed over a subtle, solid brand background.leerob5m ago@cursor can you review this PR?cursorbotreviewed 3m agosrc/vs/workbench/composer/browser/components/ComposerUnifiedDropdown.tsx3292- {selectedMode().keybinding}3293+ {composerOpenModeToggleKeybinding()}Bug: Function Returns Object Instead of String (Logic bug)The composerOpenModeToggleKeybinding is a function that needs to be called to get its value. This makes the condition always truthy.Fix in CursorFix in Web
[H2] Web & Mobile
Run cloud agents from your browser or phone.Open in browser ↗AgentDashboardTodayVercel streaming SDK functionalityAnalyzing codebase·dashboard-uiModel selector dropdownPlanning next moves·stripe-landingImplement agent window2 models·agent-uiFix send button functionality2 Files+26-12·agent-uiThis WeekFix send button functionality2 Files+26-12·agent-uiImplement user authentication3 Files+45-8·auth-systemDesign responsive layout5 Files+30-3·frontendSet up database schema4 Files+35-2·databaseCreate API endpoints6 Files+50-5·backendAsk Cursor to build, plan, fix...
[H2] Understands your codebase, no matter the size
Cursor deeply learns your codebase before writing a single line.
[H2] Multiple models
Subagents run in parallel to explore your codebase, with each one using the best model for the task.Learn more →Started 4 subagentsSet up model architectureEditing files•Opus-4.8Mission Control InterfaceBuilding dashboard•Composer 2.5Add evaluation metricsWriting tests•GPT-5.5Implement training loop with AMPPending • Gemini 3.1 Pro
[H2] Codebase indexing
A custom embedding model gives agents best-in-class recall across large codebases.Learn more →How does the payment flow handle failed transactions?
[H2] Team rules
Teach Cursor your preferences, from team conventions to specific architectural decisions.Learn more →
[H2] Spans the full development lifecycle
Cursor supports every phase from planning to writing to reviewing code.
[H2] Plan
For complex tasks, Cursor asks clarifying questions, builds a plan, then executes in the background.Learn more →How should Mission Control be opened?1Gesture (swipe up with 3 fingers)2Keyboard shortcut (e.g. CMD+F3)3Both keyboard and buttonSkipContinueAdd follow-up...Plan
[H2] Design
Visually edit any page by selecting an element to instantly rewrite, resize, or move it.Learn more →Software creation is changing. We are a group of researchers, engineers, and technologists inventing at the edge of what's useful and possible.We have much to learn, try, and build.Join our team →Shorten TaglineAgentComposer 2.5
[H2] Debug
Cursor instruments your code and uses real execution data to pinpoint the fix.Learn more →Chart tooltips freeze when hovering over data points.Checking server outputAdding console logsTaking screenshotReading ChartRenderer.tsxFound it: stale closure in the hover handler.Tooltip.tsx+3-1Fixed. Tooltips should update smoothly now.
[H2] Equipped to do real engineering
Cursor edits files, runs terminal commands, searches the web, and more.
[H2] Terminal
Run shell commands directly from Cursor, from builds to tests to installs. Sandboxed by default.Learn more →build this projectReading package.jsonRunning terminal command$ npm run buildBuild complete. Ready to deploy?
[H2] Add context
Point Cursor at exactly what matters with @-mentions and image uploads for reference.Learn more →Make drawer.tsx use vaul.emilkowal.ski and match our brandAgentComposer 2.5
[H2] Git & checkpoints
See how your code has evolved, and roll back to a previous snapshot anytime.Learn more →Set up Next.js projectJan 8Add Google OAuthJan 12Build canvas editorJan 18Add multiplayerYesterdayImprove performance3h agoAdd keyboard shortcuts1h agoShip to productionNow
[H2] Extend with tools and knowledge
Give Cursor your existing context, and add custom capabilities.
[H2] Plugins
Browse and install community-built plugins to extend Cursor with new capabilities.Learn more →/add-pluginAgentComposer 2.5Add pluginSlack Messaging KitFigma Visual EditorNotion Workspace IntegrationLinear Agent ExcellenceGitHub Actions Suite
[H2] Skills
Add domain knowledge to let Cursor discover and run specialized prompts and code.Learn more →let's /apply-notion-styleguide and /AgentComposer 2.5Skills/fix-merge-conflictsResolve git conflicts automatically/code-reviewAnalyze code for issues and improvements/apply-notion-styleguideFormat code to match team standards/prCreate a pull request with summary/testGenerate unit tests for selected code
[H2] MCP
Connect external tools and data sources like GitHub and Figma directly to Cursor.Learn more →
[H2] Trusted by the world's best developers
Watching a dozen agent branches merge every day has become normal, and that freed-up velocity shows up everywhere from release cadence to bug-backlog burn-down. Cursor isn't a convenience add-on; it's a scale-multiplier for the whole org.Cody De Arkland Senior Director, SentryI will never not be amazed to be merging PRs while riding Peloton. I'm never touching a laptop again.Nick Dobos Prompt Engineer, The Browser CompanyLeave it to the folks at Cursor to bring out the absolute most from the models. GPT-5.2 is an incredibly hard-working model, but to get it to work for a whole week is crazy. Building a whole browser is even crazier!Sherwin Wu Head of Engineering, OpenAICoding agents like Cursor have become the killer app for AI. Not only do coding agents increase the speed at which code is created, they also improve code quality.Alexis Lê-Quôc CTO & Co-Founder, DatadogCursor has helped me not only be more productive but also more curious and confident to explore new problem spaces. It's genuinely become an indispensable tool in my daily workflow.Lathesh Karkera Software Engineer, eBayMore than 70% of our engineers now use Cursor, and we've seen meaningful gains in day-to-day development, faster execution on large-scale migrations, increased rate of debugging, and even faster onboarding.James Reggio CTO, Brex
[H2] Recent highlights
Securely indexing large codebasesBy securely reusing a teammate's existing index, we cut time-to-first-query from hours to seconds on the largest repos.Research · Jan 27, 2026Scaling long-running autonomous codingWe've been experimenting with running coding agents autonomously for weeks at a time.Research · Jan 14, 2026Dynamic context discoveryAs models improve as agents, we've found success by providing fewer details up front, making it easier for the agent to pull relevant context on its own.Research · Jan 6, 2026View more posts →
[H2] Changelog
3.8Jun 18, 2026Improvements to Cursor Automations3.7Jun 17, 2026Cloud Environment Setup and Cloud Subagents in Agents WindowJun 10, 2026Bugbot is now over 3x faster, 22% cheaper, and finds 10% more bugs3.7Jun 5, 2026Design Mode ImprovementsSee what's new in Cursor →
[H2] Try Cursor now.
Download for macOS⤓Try mobile agent→

Changelog3.8 Jun 18, 2026 · ChangelogCursor Automations save you time by automating repetitive tasks with always-on agents. This release introduces the /automate skill, new triggers for GitHub and Slack, and support for computer use.
[H2] #/automate skill
Use /automate to create an automation directly in your local agent session.
Describe the task you want to automate in plain language and Cursor will configure the triggers, instructions, and tools for you.
[H2] #An emoji trigger for Slack
React to any Slack message with a designated emoji to kick off an automation. At Cursor, we use this to trigger specific automations right from Slack.
[H2] #New GitHub triggers
Automations now support five additional GitHub triggers:
Issue comment: when a comment is made on a non-PR issue
PR review comment: when an inline comment is left on a pull request diff
PR review submitted: when a PR review is submitted
Review thread updated: when a review thread on a pull request is marked resolved or unresolved
Workflow run completed: when a GitHub Actions workflow run finishes on a pull request or branch
We've added new templates for triaging failed GitHub actions and auto-fixing PR review comments to the Cursor Marketplace to help you get started.
[H2] #Computer use tool for automations
Cloud agents kicked off by automations can now use their own computers to produce demos or artifacts of their work.
The computer use tool is enabled by default for every automation, just tell the agent to include a demo of its work in your instructions.
To get started, update to the latest version of Cursor. Learn more in our docs.
[H3] Improvements (3)↓↑
Automations can now be saved in an incomplete state, so you can navigate away to set up an MCP auth without losing your progress
Automations can now open PRs by default; so you no longer have to specify that tool in the UI
You can now delete memory files in the UI, or prompt your automation to delete outdated memories when it runs
3.7 Jun 17, 2026 · ChangelogThis release introduces updates to cloud agents in the Agents Window of the Cursor desktop app.
[H2] #Cloud environment setup
Cursor can now help you set up your dev environment in the cloud in less than 10 minutes. You can watch the agent's progress in a shared terminal session as it handles setup tasks like installing dependencies.
Your environment is captured in a reusable snapshot, so future cloud agents start up faster with the ability to test changes by running your software. It can iterate over long time horizons until outputs are verified. This benefits your entire team when committed to .cursor/environment.json.
[IMG: Cloud environment setup]
[H2] #Cloud subagents with /in-cloud
Use /in-cloud to spin up a cloud subagent in its own VM to work on the next task you submit. It runs on its own VM and branch, so your local workspace stays clean and responsive.
This is especially useful for isolating long-running or parallel work like fixing CI, investigating an issue, or exploring a codebase while you keep working locally.
You can also ask a cloud subagent to babysit a PR by clicking on the quick-action pill or using /babysit. The cloud agent will iterate remotely to prepare your PR for merge without tying up the local session.
The cloud subagent can run in the background without interrupting the parent agent, which can continue to run locally or in the cloud.
[H2] #Handoff between local and cloud
Move agent sessions more reliably between your local computer and the cloud. You can offload long-running work from your machine and run as many cloud agents in parallel as you want. Pull a cloud agent back down to local to test changes yourself.
[IMG: Handoff between local and cloud]
Jun 10, 2026 · ChangelogThe average review time for Bugbot is now ~90 seconds, down from ~5 minutes. Bugbot also finds 10% more bugs per review on average — 0.62, up from 0.56 — and costs ~22% less per run.
[IMG: Bugbot is now over 3x faster, 22% cheaper, and finds 10% more bugs per review.]
[IMG: Bugbot is now over 3x faster, 22% cheaper, and finds 10% more bugs per review.]
These performance gains are made possible by progress we've made training Composer 2.5, which now powers Bugbot. Bugbot respects model block lists, and speed and performance can vary depending on your configuration.
[H2] #Run Bugbot before you push
You can now run Bugbot and Security Review with /review before pushing code. /review prompts you to choose which agents to run, or use /review-bugbot and /review-security directly.
/review also syncs with Bugbot on GitHub and GitLab. If you run /review and then open a PR with the same diff, Bugbot recognizes it, skips the review, and leaves a comment noting it has already reviewed that diff.
Available in Cursor 3.7+ and on cursor.com/agents, with support in CLI coming soon.
[H2] #Only review what's new in your PR
You can now configure Bugbot to only review what's new since the last review, keeping feedback focused on your latest updates.
Learn more in our docs.3.7 Jun 5, 2026 · ChangelogWith Design Mode in the Cursor browser, you can click, draw, or describe changes by voice to help agents update your UI.
[H2] #Multi-select elements
Click on two or more elements together in the browser. Cursor sees the selected elements, their code, the surrounding layout, and the visual relationships on the page.
Ask the agent to make one match the other, remove repeated content, or adjust a group of components at once.
[H2] #Voice input
Narrate changes through the Design Mode overlay. The mic stays available while an agent is mid-run, so you can queue the next change by voice without waiting for the previous one to finish.
Jun 4, 2026 · ChangelogWe've shipped a batch of new functionality across the TypeScript and Python SDKs. You can now choose how agent and run metadata is persisted, expose your own functions to the agent as tools, route local tool calls through auto-review, and nest subagents to any depth. This release also brings a set of reliability, performance, and platform fixes that make local and cloud SDK agents easier to run in production scripts, CI, and custom integrations.
[H2] #Custom tools
You can now hand the local agent your own tools by passing function definitions through local.customTools, on Agent.create() or per send(). The SDK exposes them to the agent through a built-in MCP server called custom-user-tools, so the model calls your code through the same path and the same permission gate as any other MCP tool.
Before this, exposing a custom capability meant standing up your own stdio or remote HTTP MCP server and wiring it into the agent. Now a function definition is enough. Custom tools are also visible to every subagent of a parent agent, so a tool you define once is available throughout the whole run.
[H2] #Auto-review
By default, a local SDK agent runs tool calls without asking for approval, since there's no human in the loop in a headless run. Set local.autoReview to route those calls through auto-review instead. A classifier decides which calls run automatically and which to hold back, rather than bypassing review entirely.
You steer that classifier with natural-language instructions in permissions.json. The autoRun.allow_instructions field describes call shapes to lean toward allowing, and autoRun.block_instructions describes the ones to hold for review. For example, you can allow read-only inspections of build artifacts while always pausing on destructive operations like deletes.
{
"autoRun": {
"allow_instructions": [
"Read-only inspections of build artifacts under ./dist are fine."
],
"block_instructions": [
"Always pause delete operations so I get a chance to review them."
]
}
}
[H2] #JSONL and custom stores
Both SDKs persist agent and run metadata so you can resume an agent after a process restart. Until now, that store was SQLite. You can now opt into a JSONL store instead, which writes a plain, append-only file you can read, diff, and check into version control. Both SqliteLocalAgentStore and JsonlLocalAgentStore are exported directly.
If neither default fits your setup, implement the public LocalAgentStore interface and pass it through local.store. Build an in-memory store for ephemeral CI runs, or back persistence with Postgres when you want agent state to live next to the rest of your application data. The Python SDK exposes host, JSONL, and composed JSONL stores through the bridge.
[H2] #Nested subagents
Subagents can now spawn their own subagents, and so on. A reviewer subagent can delegate to a test-writer, which can delegate further, with each level keeping its own prompt and model. There's nothing to turn on; a subagent session registers the executor it needs to call Task, so nesting works automatically for any agent that defines subagents.
[H2] #Reliability, performance, and platform improvements
This release also includes a batch of quality-of-life fixes across both SDKs.
[H3] Reliability↓↑
Run correlation: Every send() now carries a platform-generated requestId, exposed on Run and RunResult and persisted across the in-memory, SQLite, and JSONL stores. Tie a script or CI run to backend logs, analytics, and support threads without inferring it from agentId.
Reliable wait() on local runs: Local runs no longer resolve wait() before the terminal result is written. Hydration keeps refreshing until the run reaches a final state, so automation reads a complete result.
Safe checkpoints on dispose: Disposing a local agent no longer removes checkpoint data when a root reference is missing but checkpoint blobs still exist. The agent directory is only cleared when there's genuinely nothing left to keep.
Cloud streaming over HTTP/1.1: Cloud agent sessions now stream correctly on HTTP/1.1 transports used by some proxies, older Node fetch stacks, and certain CI images. HTTP/2 behavior is unchanged.
[H3] Performance and packaging↓↑
Lighter import: Importing @cursor/sdk no longer eagerly loads the full local agent stack. Cloud-only and type-only consumers skip the local runtime cost until the first local call, with no API change. The first local call pays a one-time import, then stays cached.
Self-contained TypeScript types: Published .d.ts files no longer reference unpublished workspace packages. This fixes TS2305 and TS2307 errors under skipLibCheck: false and silent any on stream types like TurnEndedUpdate.
Bundled ripgrep: Local shell runs use the bundled platform rg binary without modifying your global PATH. On Windows, prepending ripgrep no longer clobbers the Path variable.
[H3] Models↓↑
Composer 2 routes to Composer 2.5: SDK clients still pinning retired composer-2 slugs are routed to Composer 2.5 automatically, keeping fast variants intact, so older scripts keep running.
[H3] Python SDK↓↑
Workspace-scoped list_runs: Client, AsyncClient, and Agent.list_runs take an optional cwd, and the bridge falls back to its launch workspace. This fixes spurious "agent not found" results when the bridge runs as a subprocess.
Clearer not-found errors: Looking up an agent that isn't in the resolved workspace returns a clear not-found error instead of an opaque internal error.
0.1.6 release and analytics: cursor-sdk 0.1.6 documents the Buildkite release path and labels SDK usage as sdk-python- for clearer analytics.
Run npm install @cursor/sdk or pip install cursor-sdk to upgrade. Scripts pinning composer-2 move to Composer 2.5 automatically, and requestId is a safe addition to your run metadata schema. See the TypeScript and Python docs for full details.