Latest Updates

The AI development landscape changes weekly. New models drop, pricing shifts, and features land across all three tools. This page tracks the updates that matter so you can stay current without watching every changelog.

July 2026 Highlights

July reset the baseline across the stack: OpenAI launched the GPT-5.6 family and folded Codex into ChatGPT desktop, Cursor shipped 3.11 plus Grok 4.5, and Claude Code added Sonnet 5 to its account/provider-specific model routing while hardening its always-on background-agent runtime.

GPT-5.6 Sol, Terra & Luna July 9: Sol is the flagship, Terra balances capability and cost, and Luna targets high-volume work. All support max reasoning; the API adds persisted reasoning, explicit caching, Programmatic Tool Calling and multi-agent beta.

Codex joins ChatGPT desktop + CLI 0.144.1 July 9: Codex is integrated into the ChatGPT desktop app with inline editing, PR review and multi-repo projects. CLI 0.143–0.144 adds remote plugins by default, GPT-5.6 on Bedrock, writes-only approvals and interactive MCP auth.

Cursor 3.11 + Grok 4.5 July 8–10: jointly trained Grok 4.5 lands across desktop, web, iOS, CLI and SDK without replacing the smaller Composer 2.5; Cursor 3.11 adds durable side chats, transcript search, redesigned project pickers and conversation-level cloud hooks.

Claude Code 2.1.207 + Sonnet 5 Claude Sonnet 5 is now the default on standard Anthropic subscription seats; selected premium/direct-API accounts default to Opus 4.8, while managed-cloud defaults differ. Releases 2.1.185–207 make background agents durable across restarts, add MCP CLI auth, sandbox credential controls, dynamic-workflow sizing, stacked skills and a diagnostic/fix-capable /doctor.

June 2026 Highlights

The autonomous-work race kept pace into June: Cursor pushed Design Mode into the browser and canvases, Claude Code locked in ultracode as its dynamic-workflow trigger keyword, Codex shipped controller pairing for headless fleets, and Anthropic shipped Claude Fable 5, a new model tier above Opus.

Claude Code: Artifacts (beta) June 18: turn a Claude Code session into an interactive page — a PR walkthrough or living dashboard — shared at a private link. Beta on Team and Enterprise.

Cursor 3.8 — /automate June 18: describe an automation in plain language and the agent wires up the triggers, instructions, and tools. Plus a Slack emoji trigger and five new GitHub triggers.

Cursor — cloud subagents (/in-cloud, /babysit) June 17: spin up cloud subagents in isolated VMs, have them prepare PRs remotely, and set up a cloud dev environment in under 10 minutes.

SpaceX signs agreement to acquire Cursor June 16: SpaceX agrees to acquire Anysphere/Cursor for stock at an implied $60B equity value. The transaction remained pending regulatory approval and closing conditions, with Q3 2026 completion expected.

Codex v0.141 + Record & Replay June 18: encrypted Noise relay channels for remote executors and per-thread executor MCP servers; the Codex app's Record & Replay turns a demoed workflow into an editable skill.

Claude Code v2.1.172–183 Nested sub-agents up to 5 levels deep, implicit agent teams (TeamCreate/TeamDelete removed), parameter-scoped permission rules, and auto-mode guardrails against destructive git/terraform.

Claude Fable 5 — a new tier above Opus June 9: a Mythos-class model made safe for general use; in Claude Code from v2.1.170 (`/model fable`) and Cursor's model picker. $10/$50 per MTok; temporary subscription inclusion ultimately ended July 7, after which eligible paid-plan use consumes credits.

Cursor 3.7 — Canvas Design Mode June 4–5: Design Mode now works in the browser and inside canvases — multi-select elements, voice narration while an agent runs, an interactive context-usage report, and full-screen shared canvases.

Cursor SDK — custom tools & auto-review June 4: expose your own functions to the local agent via `local.customTools`, route headless tool calls through an auto-review classifier, pick JSONL/SQLite stores, and nest subagents to any depth.

Cursor 3.6 — Auto-review Run Mode May 29: a run mode that works longer with fewer prompts — allowlisted calls run immediately, sandboxable calls are sandboxed, the rest go to a classifier subagent. Plus Organizations for Enterprise (June 3).

Claude Code: `ultracode` is the workflow keyword v2.1.160 (June 2) renamed the dynamic-workflow trigger from `workflow` to `ultracode`, highlighted in violet. v2.1.156–158 added `.claude/skills` plugin auto-load and auto mode on Bedrock/Vertex/Foundry.

Claude Code v2.1.166 — fallback models June 6: configure up to three `fallbackModel`s tried in order when the primary is overloaded, glob patterns in deny rules (`*` denies all tools), and hardened cross-session messaging.

Codex v0.137 — controller pairing, parallel web search June 4: `codex remote-control` clients can initiate pairing and manage controller grants; plugin list gains JSON output; standalone web searches run in parallel. v0.136 added `/archive` sessions.

New official skills publishers OpenAI, Google Gemini, LangChain, Vercel, Prisma, Convex, Deepgram, Box, Contentful, Contentstack, and Base now publish official agent skills on skills.sh.

May 2026 Highlights

A pivotal month: Anthropic shipped Claude Opus 4.8 and dynamic workflows, and all three tools rounded out their first-class autonomous-work primitives within weeks of each other.

Claude Opus 4.8 + dynamic workflows May 28: Anthropic's new flagship — around four times less likely than Opus 4.7 to leave code flaws unflagged — plus dynamic workflows (research preview) that orchestrate tens to hundreds (up to 1,000) of subagents from a script, and `ultracode` to let Claude decide when to run one.

Claude Code: /goal and agent view `/goal` sets a verifiable end state and Claude works turn-by-turn until met. `claude agents` is the new dashboard for every background session. Both shipped in v2.1.139 on May 11.

Cursor 3.3 — PR review, Build in Parallel May 7: redesigned PR review (Reviews/Commits/Changes tabs), Build in Parallel quick action, Split changes into PRs, pinned quick-action pills. Bugbot moves to usage-based billing.

Codex v0.130 — Plugin sharing, remote-control May 8: `codex remote-control` for headless orchestration, plugin sharing with link metadata + discoverability, AWS Bedrock console-login, multi-environment `view_image`.

Codex v0.129 — `/vim`, `/hooks` browser May 7: Vim modal editing in composer, redesigned resume/fork picker, theme-aware status line, `/hooks` browser with pre/post-compaction execution.

Cursor in Microsoft Teams May 11: mention `@Cursor` in any Teams channel to delegate work to cloud agents. Automatic repo and model selection from prompt context.

Cursor 3.2 — `/multitask` April 24: async subagents in parallel rather than queued. Multi-root workspaces for cross-repo agent sessions in a single conversation.

Codex v0.128 — Persistent `/goal`, permission profiles April 30: persistent `/goal` workflows that survive --resume, built-in permission profiles, plugin marketplace install, MultiAgentV2 made explicit.

Claude Code: 5-hour limits doubled May 2026: 5-hour usage limits doubled for Pro / Max / Team / Enterprise to support the new long-running agent workflows.

April 2026

Cursor 3.0 — Agents Window and Design Mode

A fundamental shift from the VS Code fork to a purpose-built agent workspace:

Agents Window

New multi-repo interface for running parallel agents across environments — local, worktrees, cloud, and remote SSH. Centralized sidebar shows all agents. Toggle via Cmd+Shift+P → Agents Window.

Design Mode

Browser-based UI annotation for precise visual feedback. Use ⌘+Shift+D to toggle, Shift+drag to select areas, ⌘+L to add elements to chat, ⌥+click to add to input.

Agent Tabs & Worktrees

View multiple chats side-by-side or in a grid. New /worktree command creates isolated git worktrees. /best-of-n runs tasks across multiple models in parallel and compares outcomes.

Plugins & Enterprise

MCP Apps support structured content for richer outputs. Third-party plugins default to off for Enterprises. Directory group names in audit logs. Self-hosted cloud agent secret management.

Claude Code v2.1.84–92

Major feature additions across multiple rapid releases:

PowerShell tool (Windows): Native Windows development without WSL — opt-in preview for PowerShell as a first-class tool
Bedrock setup wizard: Interactive guided setup from login screen — AWS authentication, region configuration, credential verification, and model pinning
/powerup command: Interactive lessons with animated demos teaching Claude Code features
Per-model cost breakdown: /cost shows per-model and cache-hit breakdown for subscription users
Deferred hook permissions: "defer" decision in PreToolUse hooks for headless sessions, plus PermissionDenied hook firing after auto mode denials
Write tool 60% faster: Improved diff computation on large files
Named subagents: Subagents appear in @ mention typeahead with declarable initialPrompt frontmatter
Conditional hooks: if field for hooks using permission rule syntax
Transcript search: Press / in Ctrl+O mode, n/N to step through results

ChatGPT Desktop (then Codex App) & CLI Updates

Desktop Codex and CLI improvements across April releases (the desktop product was then branded Codex app):

Thread search: Search past threads with sidebar shortcut and keyboard shortcuts for jumping to recent threads
Theming: Choose a base theme and adjust accent, background, foreground colors and fonts
Revamped automations: Choose local or worktree execution, custom reasoning levels, and reusable templates
Windows sandbox: Proxy-only networking with OS-level egress rules for the Windows app
Plugins first-class: Sync product-scoped plugins at startup, browse in /plugins, clearer install/remove handling
Sub-agents: Path-based addressing (/root/agent_a) with structured inter-agent messaging and listing

March 2026

GPT-5.4 Released

The first general-purpose model with native computer-use capabilities, now the default across all OpenAI surfaces:

Native Computer Use

First general-purpose model with state-of-the-art computer-use capabilities — 75% on OSWorld. Agents can operate computers and carry out complex workflows across applications.

1M Token Context

Supports up to 1M tokens of context. Incorporates the industry-leading coding capabilities of GPT-5.3-Codex while improving tool use across software environments.

Top Benchmarks

57.7% on SWE-bench Pro (coding), 83% on GDPval (knowledge work). Significant improvements across professional tasks involving spreadsheets, presentations, and documents.

Available Everywhere

Rolling out across ChatGPT, Codex, and the API as the default model. GPT-5.4 Pro variant available for maximum performance on complex tasks.

Codex Security (Research Preview)

OpenAI launched Codex Security — an application security agent available to Enterprise, Business, and education customers:

Threat modeling: Builds project-specific threat models capturing system structure, trust boundaries, and exposure points
Deep vulnerability detection: Identifies complex vulnerabilities that static analysis tools miss, categorized by real-world impact
Proven at scale: Scanned 1.2M+ commits in beta, found 792 critical findings including GnuTLS heap overflows and 2FA bypasses
Editable threat models: Teams can edit generated threat models to keep the agent aligned with their security requirements

Cursor Automations

Always-on agents that run on schedules or respond to events:

Event triggers: Slack, Linear, GitHub, PagerDuty, and webhooks — agents spin up cloud sandboxes with configured MCPs and models
Memory tool: Agents learn from past runs and improve with repetition across recurring tasks
Security templates: Reviewing 3,000+ internal PRs per week, catching 200+ vulnerabilities with dedicated security audit agents
Incident pipeline: PagerDuty alerts → Datadog log analysis → Slack summaries with proposed fixes, fully automated

Cursor Composer 2

Cursor’s frontier coding model built in-house:

Architecture: Mixture-of-Experts (MoE) built on Kimi K2.5, enhanced with Cursor’s own continued pretraining and reinforcement learning
Benchmarks: 61.3 CursorBench, 61.7 Terminal-Bench 2.0, 73.7 SWE-bench Multilingual — beats Claude Opus 4.6 on Terminal-Bench
Pricing: $0.50/$2.50 per 1M tokens (standard), $1.50/$7.50 (fast variant with same intelligence)
200K context: Optimized for multi-file edits, code generation, refactoring, and long task chains across hundreds of actions

Cursor Self-Hosted Cloud Agents

Enterprise feature for running cloud agents within customer infrastructure:

Codebase, tool execution, build outputs, and secrets never leave the customer network
Same capabilities as Cursor-hosted agents: isolated VMs, full dev environments, multi-model harnesses, plugins
Up to 10 workers per user, 50 per team
Used by Brex, Money Forward, and Notion

ChatGPT Desktop for Windows (then Codex App) + CLI 0.117–0.118

Codex expanded platform reach and CLI capabilities:

Windows app (March 4): Native PowerShell support and Windows-native agent sandbox with OS-level restricted tokens and filesystem ACLs
CLI 0.117: Plugins as first-class workflow with product-scoped sync, sub-agents with path-based addressing, terminal title configuration via /title
CLI 0.118: Windows sandbox proxy-only networking, ChatGPT device-code sign-in, prompt+stdin for codex exec, dynamic bearer tokens for custom model providers

Claude Code v2.1.81–86

Steady stream of improvements and platform hardening:

--bare flag: Scripted -p mode skipping hooks, LSP, and plugins for clean automation
--channels: Permission relay for phone forwarding to Cursor-like mobile experiences
Managed-settings drop-in: managed-settings.d/ directory for enterprise config composition
Transcript search: Press / in Ctrl+O mode, n/N to step through matches
Ctrl+X Ctrl+E: Open external editor for prompt composition
Pasted images: Insert [Image #N] chip for reference in prompts

February 2026

Claude Opus 4.6 Released

The newest Claude model and the new default recommendation for all complex coding tasks:

Top SWE-Bench Scores

Best performance on SWE-Bench Verified among all available models. Superior agentic behavior and sustained reasoning over long tasks.

Enhanced Agentic Performance

Improved tool use across hundreds of tools. Better prompt injection resistance. More reliable multi-step task execution.

200K Context Window

200K token context with 64K output limit. Effort parameter for adjustable reasoning depth. Memory improvements for complex tasks.

Available Everywhere

Available in Claude Code, Cursor (via model picker), and via Anthropic API. Recommended with Max/Ultra subscription plans for full access.

Claude Code v2.1.30

Latest Claude Code improvements:

PDF page ranges: Read tool pages parameter to read specific page ranges from PDFs (e.g., pages: "1-5")
Token metrics: New /usage command with detailed input/output token breakdown
OAuth improvements: Better error handling and refresh token management
MCP health checks: Automatic restart for failing MCP servers

Cursor v2.4 — Subagents and Skills

Major agent experience update:

Subagents: Specialized subagents for testing, docs, refactoring. Define custom subagents in .cursor/agents/
Skills (SKILL.md): Reusable skill definitions with SKILL.md manifests. Share across projects and teams
40x faster hooks: Near-instant hook execution
Cursor Blame: AI-powered git blame with context
Agent clarification: Agents ask clarifying questions before complex tasks
CLI Plan/Ask modes: cursor --plan and cursor --ask for offline usage
Cloud handoff: Use & suffix to hand tasks to cloud agents

GPT-5.3-Codex Released

At that February release point, the model powering Codex surfaces was:

Improved coding performance across all languages
Better long-context handling with automatic compaction
Enhanced tool use for MCP integrations
Available in the desktop app (now part of ChatGPT), CLI, IDE extensions, and Cloud

January 2026

Claude Code v2.1.17-25

SKILL.md support: Define custom skills in .claude/skills/
Session forking: Create branches from any conversation point
--from-pr flag: Start with context from a GitHub pull request
Arrow key history: Navigate prompt history with up/down arrows
Cloud handoff: Use & suffix to hand off tasks to cloud agents

Cursor v2.3

Layout customization: Drag-and-drop panel arrangement
Multi-monitor support: Layout sync across displays
25% memory reduction: Better handling of large files
Improved crash recovery: Enhanced stability for long sessions

Codex Automations General Availability

Codex automations moved from beta to GA:

GitHub issue assignment to @codex
Slack integration for team channels
Linear ticket automation
Scheduled cron-based tasks
All automations run as sandboxed cloud agents

December 2025

Cursor v2.2 — Debug Mode and Visual Editing

Debug Mode

Runtime log instrumentation for automatic root cause analysis. Works across multiple tech stacks and languages.

Visual Style Editor

Real-time visual design in Cursor Browser. Modify elements and colors directly in a live preview.

Multi-Agent Judging

Run parallel agents on the same task, then automatic evaluation picks the best solution.

Pinned Chats

Pin important conversations in the agent sidebar for quick reference.

Claude Code v2.0.x — Background Agents

Background agents: Run tasks while you continue working
Named sessions: /rename to name, /resume <name> to resume
.claude/rules/ directory: Support for rules alongside CLAUDE.md
Prompt suggestions: Claude suggests follow-up prompts (Tab to accept)
Model switching (Alt+P): Switch models while composing a prompt

Codex Cloud Launch

OpenAI launched Codex Cloud — background agents running on OpenAI infrastructure:

Git worktree isolation for every task
Parallel task execution (up to 10 concurrent)
GitHub PR creation on completion
Slack and Linear notification integrations

November 2025

Claude Opus 4.5 Released

First model to break 80% on SWE-Bench Verified
Claude Code Desktop app launched (native experience with all CLI features)
Pro users gained access to Opus 4.5
Enhanced prompt injection resistance and agent capabilities

Cursor v2.0 — Multi-Agent Era

8 parallel agents: Run up to 8 agents simultaneously
Cursor Composer 1: 4x faster model (250 tokens/sec)
Sandboxed terminals: Secure execution environments (macOS)
Voice mode: Speech-to-text agent control

GPT-5.2 Codex Released

Replaced GPT-5.1-Codex-Max
Specialized for bug fixing and UI generation
24+ hour task endurance via compaction
Available in Cursor and GitHub Copilot

October 2025

Cursor v2.1 — AI Code Reviews

In-editor bug detection examining your changes
Instant Grep (beta) for all agent grep commands
Improved Plan Mode with interactive clarifying questions

Codex CLI Open-Source Release

OpenAI open-sourced the Codex CLI:

Full source available on GitHub
Community contributions accepted
Plugin architecture for extensions
Cross-platform support (macOS, Linux, Windows)

Current Model Snapshot (July 2026)

Current Recommended Models

Model	Provider	Context	Typical Use	Pricing (per 1M tokens)
Claude Fable 5	Anthropic	1M	Hardest refactors, app builds, long-running tasks	$10 / $50
Claude Opus 4.8	Anthropic	1M	Premium reasoning below Fable 5	$5 / $25
Claude Sonnet 5	Anthropic	1M	Default on standard subscription seats	$2 / $10 introductory
Claude Haiku 4.5	Anthropic	200K	Cheap parallel/bulk work, subagents	$1 / $5
GPT-5.6 Sol / Terra / Luna	OpenAI	1.05M	Frontier / balanced / efficient	$5/$30 · $2.50/$15 · $1/$6
Grok 4.5	SpaceXAI / Cursor	500K API*	Long-running coding and computer work	$2 / $6
Cursor Composer 2.5	Cursor	200K	Fast local iteration in Cursor	$0.50 / $2.50
Gemini 3.1 Pro	Google	1M	Multimodal, extreme context	$2 / $12

* A host such as Cursor may expose a smaller working context; verify the current model picker.

Previous or Superseded Guidance

GPT-4, GPT-4.1, GPT-4o (legacy relative to current GPT-5.6 guidance; availability depends on product)
GPT-5.1 / GPT-5.1-Codex / GPT-5.1-Codex-Mini/Max, GPT-5.2 (removed from the Codex model picker)
GPT-5.3-Codex, GPT-5.4, and GPT-5.5 (superseded by the GPT-5.6 family)
Cursor Composer 1 and Composer 2 (older generations; Composer 2.5 is current)
Claude Opus 4.6 / 4.7 (older than Opus 4.8; v2.1.207 moved the documented managed-cloud defaults to Opus 4.8, except Foundry on Sonnet 4.5)
Claude 3.x series (legacy relative to current 4.5+ and 5-series guidance)
o1-preview, o1-mini (obsolete)

MCP Ecosystem Growth

Recent Official MCP Servers

Atlassian MCP — Jira and Confluence integration
Cloudflare MCP — Workers, R2, and D1 management
Vercel MCP — Deployment automation
Supabase MCP — Database and auth operations
Linear MCP — Issue tracking integration
Stripe MCP — Payment processing automation

Skills Ecosystem

The Agent Skills ecosystem has expanded significantly:

Skills.sh marketplace now hosts 200+ skills
Universal CLI: npx skills add <owner/repo> works across 35+ agents
Popular skills: Cloudflare deployment, code review, Atlassian integration, API documentation generation

Security and Compliance

Enterprise Updates

SOC 2 Type II

Verify each vendor’s current trust center and the product/contract scope of its SOC reports.

Enhanced Privacy

Business/API offerings provide contractual data controls; consumer settings, retention, and training choices vary by product and plan.

Audit Logging

Enterprise audit capabilities vary by tool, surface, and contract; confirm the events covered before relying on them for compliance.

GDPR Compliance

Data-processing terms and regional residency depend on vendor and enterprise contract; verify the current DPA and region support.

How to Stay Updated

Automatic Updates

Settings > Update > Auto-update. Choose “Stable” or “Beta” channel.

claude update                 # update to latest
claude --version              # check current version

codex update                              # supported standalone installs
npm install -g @openai/codex@latest       # npm alternative
codex --version               # check current version

News Sources

Changelogs: Cursor | Claude Code | Codex
Blogs: Cursor | Anthropic | OpenAI
Social: Follow @cursor_ai, @AnthropicAI, @OpenAI on X/Twitter
Community: Join the Discord servers for real-time announcements

Breaking Changes

Recent Changes Requiring Action

Claude Sonnet 5 (June 30): default on Pro, Team Standard, and Enterprise subscription seats with native 1M context; Max, Team Premium, Enterprise pay-as-you-go, and Anthropic API sessions default to Opus 4.8, managed-cloud defaults differ, and introductory Sonnet API pricing ends August 31, 2026
Claude Fable 5: restored globally July 1; after temporary inclusion through July 7, current paid-plan use requires usage credits
GPT-5.6 family (July 9): Sol, Terra, and Luna replace GPT-5.5-era guidance; review model slugs, reasoning effort, and tier routing
Cursor Grok 4.5 (July 8): available across desktop, web, iOS, CLI, and SDK; choose standard or fast pricing
Cursor 3.0: New Agents Window interface replaces old layout — cloud agents removed from Editor
Claude Code: /simplify and /code-review are distinct commands; use /simplify for reuse/quality/efficiency review and /code-review for the dedicated review workflow
Claude Code v2.1.84+: /tag and /vim commands removed (use /config → Editor mode instead)
Claude Code v2.1.160 (June 2): the dynamic-workflow trigger keyword changed from workflow to ultracode — typing “workflow” no longer starts a run, though asking in your own words still does

Migration Notes

These rows describe documentation or workflow migrations, not blanket API deprecation or end-of-support claims. Verify product availability with the provider.

Previous guidance or workflow	Snapshot changed	Current path
GPT-5.5-era Codex guidance	July 2026	GPT-5.6 Sol / Terra / Luna
Cursor Composer 2	May 2026	Composer 2.5
Cursor v2.x Editor cloud agents	April 2026	Agents Window
Cursor Composer 1	March 2026	Composer 2 / 2.5 lineage
Cursor v1.x UI	February 2026	v3.x Agents Window
Claude Code WSL-only setup	August 2025	Native Windows or optional WSL
Older MCP specification revisions	Date-specific revisions	Current dated MCP specification and SDKs

Last updated: July 11, 2026