Why is OpenAI building a super app?

Three reasons. First, fragmentation: users juggling ChatGPT, Codex, the Atlas browser, the Sora app, the desktop app, and various plugins is bad UX and worse data flow. Second, competitive pressure: Anthropic's Claude Code has dominated the developer workflow conversation, and Google's Antigravity 2.0 unified the IDE + CLI + standalone agent. Third, agentic workflows: a single app can hold persistent memory, browser state, code state, and conversation state in one context — which is what enables 'plan a trip, book it, and put it in my calendar' or 'refactor this codebase, run the tests, and open the PR' from one prompt. Fragmented apps can't do that as well.

When will the OpenAI super app fully launch?

There is no announced single launch date as of June 21, 2026. OpenAI is rolling capabilities into the Codex desktop app progressively — computer use, built-in browser, image generation, persistent memory — and ChatGPT and Atlas will fold into that surface across H2 2026. Expect the fully unified app branding ('OpenAI desktop' or similar) by late 2026 or early 2027. Until then, current users have access to most of the components individually with deepening cross-product integration via Codex desktop, the Atlas extension, and ChatGPT memory.

How does the OpenAI super app compare to Antigravity 2.0 and Claude Code?

All three are different shapes of the same insight: agents need persistent context across surfaces. Google Antigravity 2.0 ships as a standalone desktop app + an IDE fork + a Go CLI, with Gemini 3.5 Flash powering speed. Anthropic Claude Code is terminal-first with VS Code and JetBrains integration, optimized for codebase-wide autonomous workflows. OpenAI's super app is desktop-first with a focus on consumer + prosumer workflows beyond coding — it includes Atlas browsing and image generation. Pick by primary user: Antigravity 2.0 for agent-first dev teams in Google's stack; Claude Code for terminal-native engineers on Claude; OpenAI super app for prosumers who span coding, research, content, and browsing.

Quick Answer

What Is OpenAI's Super App? ChatGPT + Codex + Atlas (2026)

Q: What is OpenAI's super app?

OpenAI's super app is a unified desktop application that consolidates ChatGPT (chat), Codex (agentic coding), and Atlas (AI browser) into one workspace. OpenAI confirmed the strategy after Wall Street Journal reporting in March 2026, and as of June 2026 the rollout is in phases — Codex desktop has been progressively absorbing browser, computer-use, image generation, and persistent memory features, laying the foundation for the full merger later in 2026. The super app is desktop-first for professionals; the standalone ChatGPT mobile app continues separately. It is OpenAI's response to the fragmentation problem and to competitive pressure from Anthropic's Claude Code and Google's Antigravity 2.0.

Published: June 21, 2026

What Is OpenAI’s Super App? ChatGPT + Codex + Atlas Unified

OpenAI is quietly assembling a desktop super app that merges ChatGPT, Codex, and the Atlas browser into one workspace. As of June 2026, the rollout is in phases — the Codex desktop app is absorbing computer-use, browsing, image generation, and persistent memory features one by one. Here’s what it is, why it matters, and how it compares to Antigravity 2.0 and Claude Code.

Last verified: June 21, 2026.

TL;DR

What it is: Unified desktop app that consolidates ChatGPT, Codex, and Atlas browser.
Status: Phased rollout via the Codex desktop app; not yet a single rebranded product.
Why now: Reduce fragmentation, enable cross-surface agentic workflows, respond to Anthropic Claude Code and Google Antigravity 2.0.
Timeline: Full unification expected by late 2026 or early 2027.
Mobile: The standalone ChatGPT mobile app remains separate for now.
Best for: Prosumers and developers who span coding, research, browsing, content creation.

The fragmentation problem

In 2024-2025, an OpenAI power user had to switch between:

ChatGPT (chat, memory, GPTs).
Codex (coding agent, CLI).
Atlas (AI browser, web actions).
Sora app (video generation).
ChatGPT desktop app (a thin chat surface).
Various plugins, custom GPTs, and connectors.

Each had its own context, its own memory, its own session state. A workflow like “research a market, write a report, build a quick prototype, and email it to the team” required hopping between four to five surfaces with manual copy-paste between them.

This is the problem the super app solves. One persistent context — chat, browser tabs, code state, memory, files — all in one workspace.

What’s been shipped so far

As of June 21, 2026, here’s what’s already been folded into the Codex desktop app:

Capability	Status (June 2026)
Agentic coding	GA
Computer use (interact with other apps)	Preview
Built-in browser	Preview (Atlas folding in)
Image generation	GA (ChatGPT Images 2.0)
Memory (Dreaming)	GA
File / repo persistence	GA
Voice mode	Coming H2 2026
Sora video generation	Folding in late 2026
Full ChatGPT chat surface	In rollout
Full Atlas browser merge	Expected Q4 2026

OpenAI’s strategy is “build the super app in the open.” Each Codex desktop update has added a new piece of what will eventually be the unified product.

How the unified workflow looks

Once the super app is fully assembled, a single prompt can chain across surfaces. Example:

You: “Research the top 5 competitors to OpenClaw, read their pricing pages, build a comparison spreadsheet, and draft an email to my team summarizing the takeaways.”

What the super app does:

Atlas browses the web, reads competitor pages, captures pricing data.
ChatGPT-style reasoning synthesizes the data into a comparison.
Codex creates a spreadsheet (or Markdown table) with the structured comparison.
Memory remembers your team’s preferences and tone for the email.
Drafts the email in your style and queues it for you to send.

All in one workspace, with one persistent context. Today this requires 4 manual handoffs across separate tools.

Why OpenAI is doing this now

Three pressures converged:

1. Anthropic’s Claude Code success. Claude Code became the dominant developer agent in 2025-2026 by going deep on a single surface (terminal) with massive context (1M tokens, Opus 4.8 dynamic workflows). It made the case that depth + persistence = winning agents.

2. Google’s Antigravity 2.0. Google shipped a standalone desktop agent + IDE fork + Go CLI at I/O 2026, all sharing a single agent harness. That architecture (one agent, multiple surfaces) is what OpenAI is now building toward.

3. Consumer + prosumer market reality. Most OpenAI revenue is consumer subscriptions, not enterprise API. The path to higher per-user revenue and stickiness is making ChatGPT users into super-app users who run their daily workflows inside OpenAI’s surface — not just chat with ChatGPT for 20 minutes a day.

How the super app fits with the ChatGPT memory “Dreaming” update

The June 4, 2026 Dreaming update is a critical piece of the super app puzzle. Without persistent, automatically curated memory across surfaces, a unified app would still feel fragmented — each “session” would start fresh.

Dreaming gives the super app a persistent profile of you that carries across chat, code, browsing, and image generation. When you ask Codex to build something, it knows your code style from your past ChatGPT conversations. When you ask ChatGPT to write an email, it knows your team and tone from your past Atlas browsing.

Comparison to competitors

Dimension	OpenAI super app	Google Antigravity 2.0	Anthropic Claude Code
Surface	Desktop (Mac, Win, Linux)	Desktop + IDE fork + CLI	Terminal + IDE plugins
Primary persona	Prosumer + developer	Developer	Developer
Includes browser	Yes (Atlas folding in)	No (separate Chrome)	No
Includes image gen	Yes (ChatGPT Images 2.0)	Limited	No
Includes persistent memory	Yes (Dreaming)	Yes (project memory)	Yes (CLAUDE.md, skills)
Best models	GPT-5.5, GPT-5.5 Instant	Gemini 3.5 Flash, 3.1 Pro	Opus 4.8, Sonnet 4.6
Multi-agent orchestration	Building (Codex parallel jobs)	Yes (multi-agent native)	Yes (Opus 4.8 dynamic workflows)
Pricing	ChatGPT Plus $20/mo, Pro $200/mo	Free tier + Google AI Pro/Ultra	Pro $20, Max tiers
Ecosystem age	Newest as super app	Newest as platform	Most mature for coding

See OpenAI super app vs Antigravity 2.0 vs Claude for the deeper breakdown.

What to do today

For most users in June 2026:

If you’re a Codex user: Install the latest Codex desktop app. You’re already on the super app path.
If you’re a ChatGPT-only user: The desktop ChatGPT app will gradually become the super app surface; no migration needed.
If you’re an Atlas user: Atlas continues; the merge into Codex desktop will be opt-in initially.
If you’re evaluating: Don’t wait for the unified launch. The pieces today are powerful enough; the unification is a UX upgrade, not a capability upgrade.

Risks and unknowns

Lock-in: A super app where memory, code, browser, and chat all live together is more switching-cost than a single chat tool. Worth thinking about.
Mobile gap: The standalone ChatGPT mobile app is separate; cross-device continuity is unclear.
Privacy implications: A single surface that knows everything about your work is also a single surface that, if breached, leaks everything. Enterprise customers should evaluate carefully.
Latency / cost: Holding more context across more surfaces costs more per request. Pricing tiers will likely change as the super app matures.

Sources

The New Stack: “OpenAI’s superapp takes shape”
Engadget: “OpenAI’s latest Codex update builds the groundwork for its upcoming super app”
PCMag: “OpenAI plans desktop superapp to combine ChatGPT, Codex, Atlas browser”
MindStudio: “OpenAI unified AI super app — ChatGPT, Codex, agentic”
eWeek: “OpenAI ChatGPT desktop superapp, Codex, Atlas, Neuron”
Wall Street Journal coverage, March 2026

Published June 21, 2026 by andrew.ooo. See OpenAI super app vs Antigravity 2.0 vs Claude and ChatGPT Dreaming memory update.