Codux is a native Rust + GPUI workspace for developers who live in AI coding CLIs. It unifies Codex, Claude Code, OpenCode and more with live agent status, token analytics, local memory, credential-safe SSH/database access, Git worktree sessions, and encrypted handoff across desktop, phone, and headless hosts.

Hey Product Hunt —I built Codux because my AI coding workflow was getting messy. I use Codex, Claude Code, OpenCode and other agent CLIs across multiple projects, terminals, Git worktrees, and machines. The tools are powerful, but the surrounding workflow was fragmented: lost context, unclear token usage, unsafe credential handling, and no good way to resume long-running sessions from another device. Codux is my attempt to build the missing control plane: a native Rust + GPUI workspace for AI coding CLIs, with live agent status, token analytics, local memory, credential-isolated SSH/database access, worktree-aware sessions, mobile handoff, and headless host support. I’d love feedback from developers who already use AI coding agents seriously. What breaks most often in your current workflow?

How does the local memory piece actually work across the different CLIs - is it one shared context that all of them read from, or do you have to set it up per agent?

@efekhlg Codux uses one local, project-aware memory layer and automatically injects the relevant context into supported CLIs, so you don’t need to set it up separately per agent.

How does the local memory piece actually work across the different CLIs - is it one shared context that all of them read from, or do you have to set it up per agent?

@efekhlg Codux uses one local, project-aware memory layer and automatically injects the relevant context into supported CLIs, so you don’t need to set it up separately per agent.

Codux

Control plane for AI coding CLIs

Control plane for AI coding CLIs