Headline: OpenAI turns Codex into a “super app” for developers — desktop agent now controls your Mac, browses pages, generates images, and connects to 90+ tools
OpenAI pushed a major update to its Codex desktop app today, consolidating a host of agentic capabilities into a single, ChatGPT-linked developer workspace. The company says more than 3 million developers use Codex weekly, and the latest release is explicitly pitched as “Codex for (almost) everything”: a multitasking, memory-enabled assistant that can interact with local apps, surf and annotate web pages, generate images, and plug into a broad swath of developer tooling.
What’s new (quick take)
- Computer control: Codex can see your screen, move the cursor, and click/type inside Mac apps. Multiple agents can run in parallel without interrupting your workflow. OpenAI highlights use cases like frontend iteration, app testing, and other workflows that don’t expose APIs.
- In-app browser and page annotations: Agents can browse and accept direct comments on pages so you can give precise instructions. OpenAI says it’s focused on frontend and game dev for now, with broader browser control to come.
- Image generation: gpt-image-1.5 is baked into the workflow — no separate API key required and usage is tied to a ChatGPT account.
- 90+ new plugins: Integrations include Atlassian Rovo, CircleCI, CodeRabbit, GitLab Issues, the Microsoft Suite, and Neon by Databricks — combining skills, app integrations, and MCP servers to extend what Codex can access and act on across a developer’s toolchain.
- Dev workflow features: multiple terminal tabs, GitHub PR review comment handling, SSH to remote devboxes (alpha), a summary pane that tracks agent plans/sources/artifacts, and rich sidebar previews for PDFs, spreadsheets, slides, and docs.
- Memory and proactive mode: Codex can learn preferences and prior actions, suggest where to start the day, and surface relevant items from Google Docs comments, Slack threads, Notion pages, and codebase context into a prioritized task list.
Why it matters for Web3 and crypto developers
- Tight CI/CD and repo integrations (CircleCI, GitLab, GitHub PR handling) can speed iteration and testing for smart contracts, dApps, and frontends.
- SSH to remote devboxes and multi-terminal workflows help with complex build environments common in blockchain infrastructure work.
- Built-in image generation could ease rapid NFT mockups and visual assets without juggling separate services.
- The proactive task list and memory features may make long-running, cross-repo projects and DAO coordination easier to manage.
Competition and context
- The new Codex release overlaps considerably with the viral OpenClaw agent framework, which emerged in early 2026 as a popular open-source local agent system that connects messaging, files, browsers, and shell commands. OpenClaw — created by Austrian developer Peter Steinberger — amassed roughly 60,000 GitHub stars in 72 hours and was widely compared to a “personal AI OS.”
- Steinberger joined OpenAI in February to lead personal agent work after outreach following OpenClaw’s rise; the project moved to an open-source foundation with OpenAI as a financial sponsor. Earlier, Anthropic had sent a trademark complaint over Steinberger’s original project name “Clawdbot,” a tussle that contributed to rebrands and, observers say, accelerated his move to OpenAI.
- Anthropic’s comparable product is Claude Code, a closed-source, terminal-focused coding agent that reads codebases, runs tests, edits files, and commits to GitHub. Anthropic added a computer-control feature for Claude in March as a macOS research preview for Pro and Max subscribers.
Availability and limits
- The update is rolling out today to Codex desktop users signed in with ChatGPT. Personalization and computer-control features are not yet available in the EU or UK.
Bottom line
OpenAI is packaging local computer control, browsing, image generation, and deep integrations into a single desktop Codex app, positioning it as a generalist developer assistant. For crypto builders juggling smart contracts, CI, remote infra, and frontend UX, the platform’s richer integrations and proactive memory could shave substantial time off iterative cycles — while raising questions about privacy, permissions, and regional availability that teams will want to evaluate.
Read more AI-generated news on: undefined/news
