opencode + tmux = get shit done in parallel

I stumbled upon opencode a few months ago, and it fundamentally changed my development workflow.

Opencode

Over the past 2-3 months, I’ve baptized a couple of my co-workers with opencode as their go-to CLI AI agent tool of choice. My junior engineer, who is not the best at CLI tools, apparently swears by it to get his work done.

Couple of us in my team struggled with VSCode’s built-in chat (maybe it has gotten better now?). It didn’t sit right with a few of us folks. Personally, I never felt like using it on a day-to-day basis. We tried out the custom chat modes and stuff, but it never clicked. It fell short in keeping track of changes on a large codebase that required so many changes at once, and navigating what changes were relevant versus irrelevant was frustrating. Plus, there was no undo button for the last prompt. I can’t remember how many times I’ve abused the /undo command in opencode.

The multiplexer-CLI agent combo

We’re faster with our agent-augmented development. And I’ve gotten creative with how I use opencode and tmux. I run multiple windows in a session across three different projects - Android, Frontend, and Backend. I parallelize the work on a feature and Claude just works in three separate project contexts while I sit back and monitor the progress, prompting away with voice.

I don’t even wanna type anymore. I thought of buying one of those split ergo keyboards. But instead I bought a DJI mic mini to whisper words to my computer, when I saw this one guy on the internet using a $10 mic to prompt AI agents.

This is the peak amount of lazy I could get. It’s even funnier that I asked Claude to edit my tmux conf to handle high throughput window switching with ctrl+tab, one less key press to get me to go faster. My coworker joked that Claude should just call when it’s done and ask for next steps – and honestly, I wouldn’t say no.