// FOR CLAUDE CODE & CURSOR

Coding got faster.
Managing agents didn't.

Parallel agents are table stakes. The bottleneck is review — knowing what shipped and why. One dashboard to watch them, answer prompts, read the diff, and merge.

Already running agents in parallel on your Mac? Same branches. Better place to catch up.

$ npm install -g octomux && octomux start
macOS · free & open source · runs on your laptop
// writing got faster · reviewing didn’t
--
stars on GitHub
--
contributors
MIT
free forever

More agents isn’t the upgrade.
Owning the merge is.

Tools made it easy to spin up parallel work. They didn’t make it easy to review five diffs before lunch.

Review before you ship

Agents implement. You decide what merges. Read the diff in-app — not a stack of approve buttons on code you never opened.

One inbox, not ten tabs

Every “allow this?” lands in one place. See what’s running vs waiting on you without hunting terminals.

Delegate work, not judgment

Agents can start the next task. Claude or Cursor per job. You still ship when you’ve actually looked.

Start a task. Watch it run. Ship it.

Agentic engineering: AI writes, you review and own the merge. Same flow for one agent or five.

01 — Start

Describe the task. Pick Claude or Cursor.

Type what you want, paste a ticket link, or open a draft to tweak the prompt first. Each task gets its own branch so agents don’t step on each other.

Home screen with the task composer and sessions inbox
Live task view with agent editing files and writing a summary
02 — Watch

One place to see everything.

Live output per task. Catch up without becoming the person who only clicks “approve” on diffs they never read.

03 — Ship

Review the diff. Open the PR.

Read the changes in the app, leave comments if you need to, hit Ship. No tab-hopping to GitHub mid-review.

Diff review with file tree and side-by-side diff

Common questions.

I already run parallel agents on my Mac — why try octomux?

Starting agents is solved. Catching up isn’t. octomux is one dashboard for what happens after they run: permissions in an inbox, status on a board, diffs to read before you ship. Plus Claude or Cursor per task, and agents that can queue the next job.

Is this for vibe coding?

No. Vibe coding is fine for throwaway prototypes. octomux is for when the code has to be correct — you review the diff, you own the merge, you can explain what shipped.

Does it work with Cursor?

Yes. Pick Claude Code or Cursor for each task. You can add a second agent on the same branch to review the diff.

What if two agents need the same file?

Each task gets its own branch, so they don’t collide on your main code. Five agents can work on the same filename in parallel — on separate branches.

Does my code leave my laptop?

No. Everything runs locally. No cloud sync or telemetry.

What if my laptop reboots?

Run octomux start. Your tasks and review progress come back.

Can I steer an agent mid-task?

Yes — send a message from the dashboard and it course-corrects without starting over.

Is it free?

Yes. MIT license. You only pay for Claude Code or Cursor.

Close your laptop.
Come back tomorrow.

Agents keep working. Your review queue is still here. Run octomux start to pick up where you left off.

$ npm install -g octomux && octomux start