Version History & Features

This comprehensive changelog tracks Cursor’s evolution, highlighting major features, improvements, and fixes across all versions. Use this as a reference for understanding feature availability and planning upgrades.

Version 3.x Series (2026)

June 2026 Updates

Latest

Cursor 3.7 — Design Mode in the Browser & Canvases (June 4–5, 2026)

Design Mode improvements (June 5)

Multi-select elements — click two or more elements in the Cursor browser and the agent sees their code, the surrounding layout, and the visual relationships between them; ask it to make one match another, remove repeated content, or adjust a group of components at once
Voice input — narrate changes through the Design Mode overlay; the mic stays live while an agent is mid-run, so you can queue the next change by voice without waiting for the previous one to finish

Canvas Design Mode & Context Usage Report (June 4)

Design Mode in canvases — select and annotate UI elements directly inside a canvas
Context usage report — view an agent’s context usage as an interactive report
Shared canvases can now open full-screen in the browser, and agents can embed clickable buttons in canvases
Better canvas type-error fixing, improved component styling, and more chart-customization options

Cursor SDK — custom tools, auto-review & custom stores (June 4)

Custom tools — hand the local agent your own functions via local.customTools; the SDK exposes them through a built-in custom-user-tools MCP server, so they pass the same permission gate as any MCP tool and are visible to every subagent
Auto-review — set local.autoReview to route a headless agent’s tool calls through a classifier (steered by natural-language autoRun.allow_instructions / block_instructions in permissions.json) instead of bypassing review entirely
Custom stores — choose in-memory, SQLite, or an append-only JSONL store for agent and run metadata; nested subagents can now spawn to any depth
SDK clients still pinning the retired composer-2 slug are routed to Composer 2.5

Cursor 3.6 — Auto-review Run Mode & Enterprise Organizations (May 29 – June 3, 2026)

Auto-review Run Mode (May 29)

A new run mode that lets Cursor work longer with fewer approval prompts and safer execution
Applies to Shell, MCP, and Fetch calls: allowlisted calls run immediately, sandboxable calls run in the sandbox, and everything else goes to a classifier subagent that allows the call, tries a different approach, or asks for your approval
Configured under Settings > Cursor Settings > Agents > Run Mode, with custom classifier instructions

Organizations for Cursor Enterprise (June 3)

New Organizations > Teams > Groups hierarchy: an Organization is the top-level container for identity, administration, and membership; Teams are operating units (department, region, subsidiary); Groups are lightweight collections that sit across or within teams
Multi-team membership, organization-level IDP management, and org-wide usage analytics with per-team drill-down
Admins can move users between teams via the dashboard, API, or CSV; new joiners inherit team settings and permissions automatically

May 2026 Updates

Cursor 3.5 — Shared Canvases, Automations in the Agents Window (May 20, 2026)

Shared canvases

Share canvases created by agents with your team (Pro, Teams, Enterprise)

/loop skill

Agents can run a prompt repeatedly on a local schedule until a target outcome is reached

Automations, reworked

Automations now live in the Agents Window
Multi-repo automations — one agent works across several codebases to deliver, test, and verify
No-repo automations for monitoring tools outside a codebase, with five new marketplace templates (Slack digest, analytics, FAQ, finance, customer-health)
Promotional pricing: 50% off agent runs for newly created automations for their first 7 days

Cursor in Jira (May 19, 2026)

Assign work to Cursor directly from Jira, or mention @Cursor on a ticket to spin up a cloud agent
Repository and model are inferred from the ticket; completed runs post a pull-request link back to the issue

Composer 2.5 (May 18, 2026)

A substantial step up over Composer 2 in intelligence and behavior — better at sustained work on long-running tasks and more reliable at following complex instructions
Pricing: Standard $0.50/M input, $2.50/M output · Fast (default) $3.00/M input, $15.00/M output
Double usage for the first week after launch

Cursor 3.4 — Full-Screen Tabs, Compact Responses (May 13, 2026)

Full-screen tabs for focused work, with a floating prompt bar
Compact chat responses with adjustable tool-call density
Eight quality-of-life improvements and nine bug fixes for performance and reliability

Microsoft Teams + Bugbot Effort Levels (May 11, 2026)

Cursor in Microsoft Teams

Mention @Cursor in any Teams channel to delegate work to cloud agents
Automatic repository and model selection inferred from your prompt and recent activity
Replies stream back into the same thread — no context switch out of Teams

Bugbot Effort Levels

Three configurable profiles: Default (efficiency), High (extended reasoning), Custom (natural-language tuning)
Default effort: 0.7 bugs found per run · High effort: 0.95 bugs found per run
Custom requires usage-based billing
Bugbot transitions to purely usage-based billing — no more seat fees; Teams bills from on-demand spend, Individuals bill from included usage

Version 3.3 - PR Review, Build in Parallel, Quick Actions (May 2026)

Released: May 7, 2026 (3.3.0) · May 6, 2026 (3.3 preview)

The agent-execution-runtime story matures into shippable PR workflows.

Major Features

Redesigned PR Review Experience

Three tabs: Reviews, Commits, Changes
Inline review threads and top-level PR comments now visible without leaving Cursor
File-tree navigation for large pull requests
Better diff context with multi-line snippets

Build in Parallel from Plans

New quick action: click Build in Parallel on any plan
Cursor identifies independent steps and runs them simultaneously via async subagents
Dependent steps stay properly ordered — Cursor keeps the DAG intact
Faster from plan → executing change set

Split Changes into PRs

New quick action splits a single working set into multiple logical PRs
Uses chat context to identify slices; defaults to independent PRs unless dependencies require chaining
Creates a backup snapshot before proposing the split plan for approval

Quick-Action Pills

Pin your most-used skills as pinned pills above the chat input for one-click invocation
Persist across sessions and projects

Productivity Improvements

/multitask now exposes Explore-subagent controls (max depth, concurrency, cost ceiling)
Context-usage breakdown shows what is consuming context across rules, skills, MCPs, and subagents (released May 6, 2026)
Enhanced MCP connection stability under high parallelism
Better long-running job monitoring in the Agents Window status bar

Enterprise Updates (May 4, 2026)

Granular model access controls: provider-level and model-level blocklists per role/group
Soft spend limits with alerts at 50% / 80% / 100% thresholds (no hard cut-off)
Usage analytics filterable by user and product surface (chat, agent, cloud, CLI)

Version 3.2 - Agent Execution Runtime (April 2026)

Released: April 24, 2026

Cursor reframes the IDE as an agent execution runtime — the agent is the primary process, the editor is the visualization layer.

Major Features

/multitask Command

Cursor spawns async subagents in parallel rather than serializing requests in a queue
Cursor automatically breaks larger tasks into smaller chunks and assigns each to its own subagent
Use when a task naturally decomposes — multi-file refactors, parallel test fixes, cross-module updates

Multi-Root Workspaces

A single agent session can target a workspace composed of multiple folders
Cross-repository changes (frontend, backend, shared libraries) inside one cohesive session
Persists across reopens; integrates with the Agents Window

Notable Improvements

Better subagent failure recovery and resume logic
Plan-mode improvements for branched plans
Tighter integration between /worktree, /best-of-n, and /multitask — they can now share their underlying worktree pool

Version 3.1 - Tiled Layout & Voice Upgrade (April 2026)

Released: April 13, 2026

See the official Cursor changelog for the latest updates.

Major Features

Tiled Layout for Parallel Agents

Split your current view into panes to run and manage several agents in parallel
Expand panes, drag agents between tiles, persistent layouts across sessions
Ideal companion to the v3.0 Agents Window

Upgraded Voice Input

Records the full voice clip and transcribes with batch STT for higher-quality speech-to-text
Activate with Ctrl+M
New UI controls — waveforms and timers

Branch Selection Before Cloud Agents

Search and select a specific branch before launching cloud agents
Fewer manual switching steps when orchestrating long-running jobs

Productivity Improvements

Diff → file navigation for jumping to exact lines with full editor capabilities
Include/exclude filters for “Search in Files”
Enhanced prompt button stability during voice processing
Improved plan tab document behaviors
macOS text anti-aliasing refinements
Design Mode keyboard navigation improvements

April Mid-Release Updates

Interactive Canvases (April 15, 2026)

New canvas surface for dashboards, tables, diagrams, and charts
Blend live data with AI-generated structure in a single pane
Useful for spec exploration and quick visual reports

CLI Debug & Config (April 14, 2026)

New /debug CLI command for runtime diagnostics
New /btw for side questions without derailing the main conversation
New /config settings panel inside the CLI
Custom status bars for the CLI TUI

Version 3.0 - Agents Window & Design Mode (April 2026)

Released: April 2, 2026

Biggest shift since 1.0 — Cursor is now agent-first. The file tree is a fallback; the default view is a window of running agents.

Major Features

Agents Window

Parallel agents across local repos, git worktrees, cloud infrastructure, and remote SSH
Open with Cmd+Shift+P → Agents Window
Toggle between the Agents Window and traditional IDE mode at will
Cloud agents no longer live in the editor surface — they run inside the window

Design Mode

Annotate and target UI elements directly in the browser
⌘+Shift+D — toggle Design Mode
Shift + drag — select specific areas
⌘+L — add the selected element to chat
⌥ + click — add element to input

Agent Tabs in the Editor

View multiple chats in side-by-side or grid layouts
Works inside the Agents Window or the traditional editor

New Commands

/worktree — spin up an isolated git worktree for the agent’s changes
/best-of-n — run the same task in parallel across multiple models in separate worktrees, then compare outcomes

Notable Improvements

New Await tool — agents can wait for background shell commands or specific outputs
Large-file diff rendering performance significantly improved
Browser automation gains a screenshot-based coordinate-click fallback
Plan transcripts included in shared chats
Explorer subagent startup time optimized via caching

Enterprise Updates

Audit logs display directory group names for readability
Team-level admin controls for cloud-agent secret management
Enterprise-wide “Made with Cursor” attribution toggles

Migration from v2.x

Cloud agents removed from the Editor surface → they live in the Agents Window
Previous worktree selection deprecated → use the /worktree command

Version 2.x Series (2025-2026)

Version 2.5 - Plugins & Long-Running Agents (February 2026)

Released: February 17, 2026

See the official Cursor changelog for the latest updates.

Major Features

Plugins & Cursor Marketplace

Plugins system — extend Cursor with third-party integrations
Cursor Marketplace with launch partners: Amplitude, AWS, Figma, Linear, Stripe
Plugins run as MCP servers with sandbox access controls
Install and manage plugins from the Marketplace tab

Long-Running Agents (Research Preview)

Autonomous agents that run for hours in the cloud at cursor.com/agents
Create branches, run CI, iterate on code reviews autonomously
Available separately from the IDE (Feb 12, 2026)

Async Subagents with Nesting

Launch subagents that can themselves spawn further subagents
Deep task decomposition for complex features
Improved parallelization of independent workstreams

Usage Pool Changes

Two separate usage pools: Auto/Composer pool and API pool
Boosted limits for paying customers
Clearer usage tracking and billing

CLI Improvements (February 18, 2026)

Plan mode persistent menus for better navigation
Mermaid ASCII diagrams rendered in terminal
Ctrl+O toggle for plan mode visibility

v2.5.20 (February 20, 2026)

Minor stability fixes and performance improvements

Version 2.4 - Subagents & Skills (January 2026)

Released: January 22, 2026

See the official Cursor changelog for the latest updates.

Major Features

Subagents

Specialized subagents for specific tasks (testing, docs, refactoring)
Create custom subagents with .cursor/agents/ directory
Agent-to-agent communication and task delegation
Better parallelization with agent pooling

Skills (SKILL.md)

Define reusable skills with SKILL.md manifest files
Skills can specify tools, context, and prompts
Share skills across projects and teams
Built-in skill marketplace integration

Image Generation

AI-generated images directly in chat
Generate mockups, diagrams, and UI previews
Export to common formats (PNG, SVG)
Integration with visual editing mode

Cursor Blame

AI-powered git blame with context
Understand why changes were made, not just who made them
Link changes to issues, PRs, and discussions
Timeline view of code evolution

Agent Clarification

Agents ask clarifying questions before executing complex tasks
Configurable clarification threshold
Better understanding of user intent
Reduces back-and-forth iterations

40x Faster Hooks

Massive performance improvement for hooks execution
Near-instant hook responses
Better support for complex hook chains
Reduced latency in hook-heavy workflows

CLI Plan/Ask Modes (January 16, 2026)

Plan mode in CLI: cursor --plan "description" for offline planning
Ask mode in CLI: cursor --ask "question" for quick queries
Cloud handoff: Use & suffix to hand off tasks to cloud agents
Improved offline support: Better caching and local model fallbacks

Additional Improvements

Layout persistence across sessions
Custom keyboard shortcut profiles
Improved model switching UI
Better error recovery in multi-agent mode
Enhanced MCP server management

Version 2.3 - Layout & Stability (December 2025)

Released: December 22, 2025

Major Features

Layout Customization

Fully customizable layouts with drag-and-drop panels
Save and restore layout presets
Multi-monitor support with layout sync
Compact mode for smaller screens

Stability Improvements

Reduced memory usage by 25%
Better handling of large files (>10MB)
Improved crash recovery
Enhanced stability for long sessions

Additional Improvements

Improved diff view with inline comments
Better keyboard navigation in agent sidebar
Enhanced search filtering options
Fixed various UI rendering issues

Version 2.2 - Debug Mode & Visual Editing (December 2025)

Released: December 10, 2025

See the official Cursor changelog for the latest updates.

Major Features

Debug Mode

Runtime log instrumentation for automatic root cause analysis
Works across multiple tech stacks and languages
Model-agnostic debugging assistance
Intelligent error tracing and resolution

Browser Layout and Style Editor

Real-time visual design capabilities in Cursor Browser
Simultaneously work on code and design
Modify elements and adjust colors directly
Experiment with CSS and apply changes via agent

Plan Mode Improvements

Inline Mermaid diagrams with automatic generation and streaming
Delegate selected tasks to new agents
Enhanced control over plan execution
Better task breakdown visualization

Multi-Agent Judging

Automatic evaluation after parallel agent execution
Recommendations for best solution selection
Explanatory comments on selected approaches
Improved multi-agent coordination

Pinned Chats

Pin important chats in agent sidebar
Quick access to reference conversations
Better organization for complex projects

Additional Improvements

Browser protection controls via dashboard settings
Read-only terminal support for Ask Mode
Local parallel agent stability enhancements
Plans saved as editable disk files
System notifications for terminal and MCP actions
Keyboard shortcut improvements
AWS Bedrock performance optimization
Explorer pane auto-search functionality
Rules supporting prompts and scripts within folders

Version 2.1 - AI Code Reviews & Instant Grep (November 2025)

Released: November 21, 2025

Major Features

Improved Plan Mode

Interactive UI for answering clarifying questions during plan creation
Integrated search functionality (⌘+F) within generated plans
Better planning workflow for complex features

AI Code Reviews

In-editor bug detection examining your changes
Complements BugBot integration across GitHub, GitLab, and Enterprise Server
Proactive issue identification before commits

Instant Grep (Beta)

Instant execution of all agent grep commands
Regex and word boundary matching support
Manual codebase searches significantly faster

Additional Improvements

Enhanced terminal process awareness
Project-level hooks support
Multiple stability and performance refinements

Version 2.0 - Multi-Agents & Composer Model (October 2025)

Major Release

Released: October 29, 2025

Major Features

Multi-Agents

Eight parallel agents operating simultaneously
Isolated codebase copies using git worktrees
Remote machine support for distributed work
Coordinate complex tasks across multiple agents

Composer Model

Frontier model optimized for code generation
4x faster than similarly intelligent models
Enhanced reasoning capabilities
Better code understanding and generation

Browser (General Availability)

Full enterprise support with in-editor embedding
DOM element selection tools
Production-ready browser integration

Sandboxed Terminals (macOS)

Secure execution environment for shell commands
Configurable sandbox controls for enterprise teams
Enhanced security for automated commands

Team Commands

Centralized custom commands and rules
Distribution across team members via dashboard
Consistent workflows across teams

Voice Mode

Speech-to-text agent control
Customizable submission keywords
Hands-free coding assistance

Performance Improvements

Language Server Protocol optimization
Dynamic memory allocation for Python and TypeScript
Faster response times across the board

Plan Mode Background Processing

Separate model usage for planning and building
Parallel agent options during planning
More efficient resource utilization

Enterprise Features

Sandbox admin controls
Cloud hook distribution
Timestamped audit logging

Version 1.x Series (2025)

Version 1.7+ - Beta Features (September 2025)

Released: September 29, 2025

Major Beta Features

Cursor Browser (Beta)

In-app browser for real-time testing and verification
Test changes as you code without leaving the IDE
Verify UI changes instantly
Debug visual issues in context
Available in Beta settings tab

Native Plan Mode (Beta)

Built-in task planning before implementation
AI-assisted breakdown of complex features
Structured approach to large tasks
Plan review before execution
Available in Beta settings tab

Claude Sonnet 4.5 Support

Best coding model with 1M context window
Superior performance on coding tasks
Stronger everyday coding than the prior Opus generation
State-of-the-art results

GPT-5-Codex Model

Specialized for bug fixing
Excellent for UI generation
Fast iterations on frontend work
Available in model selection

How to Enable Beta Features

Open Settings (Cmd/Ctrl + ,)
Navigate to “Beta” tab
Enable “Cursor Browser” and “Plan Mode”
Restart Cursor if prompted

Version 1.4.x - Enhanced Agent Control (August 2025)

Released: August 6, 2025

Major Features

Agent Steerability

Improved agent control and guidance
Separate model selection per agent
Enhanced agent sidebar for management
Queued message handling

Transparency & Control

Usage and pricing visibility in chat
Context usage display for all operations
Compact chat mode for streamlined interface
Shared terminal with agent

Advanced Capabilities

GitHub support for Background Agents
Expanded read file capabilities
Enhanced directory listing tools
Improved grep and codebase search
Web search optimization

Performance

Faster Background Agent startup
Optimized edit performance
Context selection refinements

Version 1.3.x - Shared Experience (July 2025)

Released: July 29, 2025

Major Features

Shared Terminal

Direct terminal sharing with Agent
Real-time command execution visibility
Synchronized input/output

Context Awareness

Context usage display in Chat
Active Tab integration in Chat
Right-click directory send to Chat

Performance & Security

Significantly faster edit performance
Checkpoints for Notebooks
Enhanced security configurations

Version 1.2.x - Agent Intelligence (July 2025)

Released: July 3, 2025

Major Features

Agent Planning System

Structured to-do lists for complex tasks
Real-time progress tracking
Task queue management with reordering
Visible in chat and Slack integrations

Enhanced Context Management

PR indexing and semantic search
Improved code embedding model (30% better accuracy)
Git integration for commits, branches, and issues
~100ms faster Tab completions

Developer Experience

Merge conflict resolution in chat
Syntax highlighting in completions
Multiple image attachments support
VS Code 1.99 base upgrade

Patches

1.2.1-1.2.4: Performance improvements, UI polish, stability fixes

Version 1.1.x - Team Collaboration (June 2025)

Released: June 12, 2025

Major Features

Slack Integration

Launch Background Agents from Slack with @Cursor
Live updates in Slack threads
Full context understanding from conversations
Direct GitHub PR creation

Enhanced Search

In-app settings search (Cmd/Ctrl+F)
PR indexing and search capabilities
Improved deep-linking experience

MCP Improvements

Progress notifications for long tasks
Dynamic tool registration
Monorepo root context support

Patches

1.1.1-1.1.7: Workspace indexing, performance, agent reliability

Version 1.0 - General Availability (June 2025)

GA Release

Released: June 4, 2025

Major Features

BugBot Launch

Automated PR review system
Bug detection and commenting
“Fix in Cursor” integration
Smart code analysis

Memories GA

Project-specific knowledge retention
Approved memory generation
Cross-session persistence
Team sharing capabilities

Developer Tools

One-click MCP server installation
Jupyter Notebook agent support
Background Agents for all users
Enterprise restrictions to stable releases

Unified Pricing

Request-based quota system
Simplified billing model
Max Mode token pricing
Legacy mode sunset

Version 0.x Series (2024-2025)

Version 0.50.x - Architecture Refresh (May 2025)

Released: May 15, 2025

Key Changes

Unified Request Pricing

Single quota system
Token-based Max Mode
Removal of premium limits

New Tab Model

Multi-file change awareness
Syntax highlighting
Performance improvements

Background Agent Preview

Autonomous task execution
VM-based isolation
Early access rollout

Version 0.49.x - Rules & Control (April 2025)

Released: April 15, 2025

Major Features

Rules Generation

/Generate Cursor Rules command
Auto-attached rule patterns
Always-attached persistence
Agent rule editing

Enhanced Control

Terminal command approval
MCP tool permissions
Review diffs in chat
Chat history in palette

Version 0.48.x - Agent Unification (March 2025)

Released: March 20, 2025

Paradigm Shift

Unified Agent Mode

Single AI interface
Adaptive capabilities
Web search integration
Simplified UX

UI Refresh Phase 1

New default themes
Cleaner layouts
Improved @-context menu

Version 0.47.x - Stability Focus (March 2025)

Released: March 11, 2025

Improvements

Performance

Major memory optimizations
Stability improvements
Reduced crash frequency

Customization

Keyboard shortcuts panel
Early access opt-in
Auto-model selection

Version 0.46.x - Agent Evolution (February 2025)

Released: February 6, 2025

Features
MCP Updates

Agent as default interface - UI refresh (new themes) - Web search integration - Improved ignore files - 25 tool-call limit handling

Version 0.45.x - Repository Rules (January 2025)

Released: January 23, 2025

Major Features

.cursor/rules Directory

Multiple rule files
Automatic selection
Project guidelines
Team standards

Improved Understanding

Better codebase model
“Fusion” Tab model
Long context mode
Recent changes awareness

Version 0.44.x - Agent Capabilities (December 2024)

Released: December 17, 2024

Enhancements

Agent Terminal

Exit code capture
Background execution (YOLO)
Editable commands
Auto-save changes

Error Awareness

Lint error reading
Compiler error fixes
Smarter suggestions

Earlier Versions (2023-2024)

Notable Milestones

July 2023 - First public nightly build

Experimental Agents debut - Documentation features - Public changelog begins

October 2023 - VS Code 1.83 base

Major stability improvements - Bash mode experimental - SSH compatibility fixes

November 2023 - Chat improvements

Better contextual responses - Follow-up handling - UI enhancements

August 2024 - Composer evolution

New chat UX - Default-on Composer - Performance gains

Feature Evolution Timeline

AI Models

2023

GPT-3.5/4 foundation - Basic completions - Simple chat interface

2024

Claude integration - Multi-model support - Context improvements

2025

Claude Opus 4.1 - Claude Sonnet 4.5 - GPT-5 / GPT-5-Codex - Gemini 3 Pro

2026

Claude Opus 4.8 - Claude Sonnet 4.6 - GPT-5.5 (Codex default) - Gemini 3.1 Pro - Composer 2.5 - Plugins & Marketplace - Long-running cloud agents

Core Features

Autocomplete Evolution

Basic suggestions → Tab model → Multi-file awareness → Fusion model → Composer Model

Chat/Agent Progress

Simple chat → Composer → Unified Agent → Background Agents → Multi-Agents (8 parallel) → Subagents with Skills → Plugins & Marketplace → Long-Running Cloud Agents

Context Management

File references → Semantic search → PR indexing → Memories → AI Code Reviews

Integration Timeline

Git basics → GitHub integration → MCP ecosystem → Slack collaboration → Team Commands

Debugging Evolution

Manual debugging → Error awareness → Debug Mode (runtime instrumentation)

Version Patterns

Major Version Indicators

2.x - Advanced multi-agent, enterprise features 1.x - Production ready, GA features 0.5x - Architecture changes 0.4x - Feature maturity 0.3x - Rapid iteration 0.2x - Early development

Patch Patterns

.1-.3: Critical fixes
.4-.7: Stability improvements
.8-.11: Feature refinements
.12+: Extended support

Breaking Changes History

Major Breaking Changes

Version 2.0: Multi-agent architecture introduction
Version 1.0: Legacy MAX modes removed
Version 0.50: Unified pricing model
Version 0.46: Agent mode default
Version 0.45: Rules system overhaul

Migration Requirements

Always check for:

API compatibility changes
Extension updates needed
Configuration migrations
Team setting impacts

Looking Ahead

Based on the changelog patterns, expect:

Growing plugin ecosystem via Cursor Marketplace
Long-running agents becoming generally available
Enhanced subagent coordination and deeper nesting
Deeper enterprise integrations with SSO and audit logging
Expanded MCP ecosystem with official partnerships
More sophisticated plan mode with dependency graphs

Next Steps

Understanding the changelog helps you make informed decisions about updates. Continue to Upgrade Guides for safe update procedures or explore Feature Adoption strategies.