Category
Other
Which variant should this target?
Both
What would you like?
I would like HolyClaude to provide Codex chat with the same effective workspace, shell, and permission model as Claude Code, or a documented optional mode that enables near-parity.
Today, Claude Code is usable as a full coding agent inside the container, while Codex chat appears to run behind a stricter sandbox/orchestration layer. In practice this means Codex often cannot
reliably perform basic read-only inspection or standard shell actions without extra escalation, even in the same HolyClaude environment.
Requested improvement:
- Make Codex use the same trust/permission model as Claude Code when running inside a trusted HolyClaude container
- Or add an explicit setting for Codex permission parity with Claude Code
- Ensure standard read-only shell commands and workspace inspection work reliably
- Document any remaining capability differences clearly
Use case
I use HolyClaude as a self-hosted AI coding environment and want to switch between Claude Code and Codex depending on the task.
Right now, Claude Code can operate normally, but Codex is much more constrained, so the comparison is not meaningful and Codex becomes far less useful for real coding work. Even when both assistants
are running against the same mounted workspace, Codex spends too much time blocked by sandbox restrictions instead of acting like a full coding agent.
The main use case is simple: if HolyClaude supports multiple coding agents in the same container, they should have comparable filesystem and shell capabilities unless the user explicitly chooses a
stricter mode
Alternatives considered
No response
Category
Other
Which variant should this target?
Both
What would you like?
I would like HolyClaude to provide Codex chat with the same effective workspace, shell, and permission model as Claude Code, or a documented optional mode that enables near-parity.
Today, Claude Code is usable as a full coding agent inside the container, while Codex chat appears to run behind a stricter sandbox/orchestration layer. In practice this means Codex often cannot
reliably perform basic read-only inspection or standard shell actions without extra escalation, even in the same HolyClaude environment.
Requested improvement:
Use case
I use HolyClaude as a self-hosted AI coding environment and want to switch between Claude Code and Codex depending on the task.
Right now, Claude Code can operate normally, but Codex is much more constrained, so the comparison is not meaningful and Codex becomes far less useful for real coding work. Even when both assistants
are running against the same mounted workspace, Codex spends too much time blocked by sandbox restrictions instead of acting like a full coding agent.
The main use case is simple: if HolyClaude supports multiple coding agents in the same container, they should have comparable filesystem and shell capabilities unless the user explicitly chooses a
stricter mode
Alternatives considered
No response