[Feature Request] Improve file context handling and UX (Comparison with Cursor)
Resolved 💬 2 comments Opened Feb 9, 2026 by InverseDa Closed May 16, 2026
What version of the IDE extension are you using?
v0.4.71
What subscription do you have?
I feel that the current file reading and context management experience in Codex is lagging behind Cursor. I would like to propose a few improvements to streamline the workflow.
Which IDE are you using?
VS Code
What platform is your computer?
Windows 11 x86_64
What issue are you seeing?
I feel that the current file reading and context management experience in Codex is not as smooth as Cursor. There are three main friction points:
- No File Autocomplete: When typing a filename in the input box, there is no fuzzy search or tab-completion. In Cursor, I can just type part of a name and hit tab.
- Redundant Drag-and-Drop Steps: After dragging a file in, Codex often still prompts me to "provide the file" or confirm reading, instead of automatically parsing the file I just explicitly dropped.
- UI Placement: Dragged files sit in a separate "floating" area above the input box. In Cursor, files are embedded inline with the text (as chips), which feels much more integrated.
What steps can reproduce the bug?
- Open the chat input.
- Type the beginning of an existing filename (observe that no autocomplete suggestions appear).
- Drag a file from the file tree into the chat.
- Observe that the file sits above the input box rather than inline with the text.
- Attempt to ask a question about the file immediately (observe that Codex sometimes still asks for the file to be provided/read).
What is the expected behavior?
- File Autocomplete: When typing a filename (e.g.,
TS_Com), the editor should offer fuzzy search suggestions that I can autocomplete withTab(TS_Command.ts). - Seamless Drag-and-Drop: Dropping a file should instantly add it to the context and mark it as "read". I should not need to manually click to "provide" a file I just explicitly dropped.
- Inline UI: Dragged files should appear as inline blocks/chips within the text prompt (similar to Cursor's design) rather than a separate floating list above the input. This allows for more precise context referencing.
Additional information
_No response_
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗