Repository navigation
Feature request: unified Work/Codex entry point with full ChatGPT Project access #32130
Description
What feature would you like to see?
I would like the merged ChatGPT + Codex desktop app to provide one unified task entry point instead of requiring users to choose between Chat, Work, and Codex before starting.
The app should automatically route each task to the appropriate capabilities while keeping permission and runtime boundaries internal. When a task needs elevated capabilities—such as terminal execution, filesystem changes, network access, or connected-app actions—the app should clearly preview those capabilities and request approval.
The unified app should also make all existing ChatGPT Projects accessible. A user's project chats, files, instructions, sources, history, and organization should be available consistently across ChatGPT web, Work, and Codex, subject to the same account and workspace permissions. Users should not have to recreate projects or manually re-upload their context.
Local repository folders can remain distinct from cloud-synced ChatGPT Projects, but users should be able to link them when desired.
Why is this useful?
The current mode selector exposes internal product boundaries and creates unnecessary cognitive load. Many real tasks span research, documents, spreadsheets, presentations, and software development, so users may not know which mode to select.
Separately, having ChatGPT Projects on the web but not fully accessible in the merged desktop app fragments context and undermines the promise of a unified application.
A single project library and task composer could preserve specialized experiences contextually:
- Research citations and finished artifacts for knowledge work
- Diffs, branches, and test results for coding work
- Explicit approvals for sensitive capabilities
- Existing workspace and administrator controls
This would make the product feel like one assistant without weakening its security boundaries.
Related issues
The missing ChatGPT Projects portion is also reported in:
- Not updating existing projects from ChatGPT #31847
- ChatGPT Projects visible on chatgpt.com but missing from desktop sidebar after ChatGPT/Codex merge (macOS 26.707.30751) #31878
- [Windows][ChatGPT Desktop] Cloud ChatGPT projects are missing from the Projects view #32102
This request is broader: it asks for a unified task entry point and project model across Work and Codex, rather than only fixing missing projects in the current desktop sidebar.
Feedback reference
Feedback ID: 019f4b8e-234c-71e2-82d1-f9a04935cad3
Activity
github-actions commentedon Jul 10, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
igneous-prose commentedon Jul 19, 2026
I built an isolated macOS prototype exploring one possible interface for this broader unified Work/Codex request.
Video and screenshots:
The prototype places the existing authenticated Codex and ChatGPT/Work interfaces in one native workspace and demonstrates:
This is not intended to address the entire proposal in this issue, particularly unified task routing or complete ChatGPT Project integration. It is a working interaction prototype showing how the two existing surfaces could remain distinct while being usable together in one window.
The prototype uses separate local profiles and a separate bundle identity. I am not distributing modified OpenAI application code, authentication data or packaged account information.
ricso commentedon Oct 16, 2026
I strongly support this request, and I believe the underlying problem goes beyond usability or interface design.
The fundamental issue is that users cannot reasonably be expected to know, at the beginning of a project, which tools they will eventually need.
Real-world projects rarely fall neatly into predefined categories such as research, conversation, automation, or software development. They evolve organically. An idea may start as a question, develop into research, require experimentation, and eventually involve coding, hardware integration, automation, or all of these.
I have several ongoing projects that demonstrate this. One recent example started with a simple question about repurposing an Amazon Echo Show as a Home Assistant device. The discussion evolved into hardware research, purchasing decisions, physical device identification, firmware analysis, USB/Fastboot diagnostics, and ultimately plans for custom firmware, audio integration, and software development.
At no point in the beginning would it have been reasonable to classify this as a "Codex project." Yet that is where parts of the implementation may eventually belong.
Having to choose between ChatGPT, Work, and Codex upfront — or migrate between them later — creates an artificial constraint on how projects develop.
More importantly, I believe this fragmentation actively discourages exploration and creativity. Instead of freely pursuing an idea wherever it leads, users must consider which environment they're using, whether their accumulated context will survive, and whether moving forward will require rebuilding work already completed elsewhere.
I understand the value of specialized execution environments. Codex may need repository access, terminal execution, and development-specific controls, while other tasks require research, applications, or general reasoning.
But these should be capabilities made available as the project requires them, not separate environments that force users to anticipate the nature of their work.
My preferred model would be:
One project, one persistent knowledge base, one continuous conversation history, with specialized capabilities activated when needed and subject to appropriate permissions.
Users should be free to explore, experiment, change direction, and develop ideas without having to manage the boundaries between OpenAI's tools.
I believe this is not merely a convenience improvement. It is an important prerequisite for making AI genuinely useful as a long-term partner in complex, evolving projects.