cmux

You will feel this immediately if your current setup is six terminal windows and a guess. Work is organised into workspaces with splits and vertical tabs, and the deciding trait in daily use is session restore: closing the lid stops being a decision, because what you had open comes back the way you left it. Attaching a workspace over SSH means the same layout follows you to a remote box.

Pick it if organising parallel sessions is the actual chore. Pick Warp if you want the terminal to answer questions itself, or stay with your current emulator if tabs were never the problem.

The concern is co-location. An in-app browser that agents can drive shares a process boundary with the workspaces holding your remote connections, and the row records no container isolation of any kind. Content fetched by an agent is untrusted input; here it arrives inside the tool rather than beside it. Nothing documented explains what the browser may reach or what it may not.

What it does right is refuse a lie of convenience: the sidebar reports branch and pull request status while the application performs no git operations itself, so a display never implies an action it did not take.

Two points. 1. The notification path uses documented terminal escape sequences, so any program that already emits them is supported without integration work, and support does not decay when an agent changes its output format. That is interoperability by standard rather than by adapter. 2. The application hosts agents and runs no loop of its own, which means model change is somebody else's migration.

It is written in Swift on an existing terminal core rather than a web shell, so its performance claims rest on architecture rather than assertion. No evaluation is published and none would be meaningful here.

Manaflow has distribution most seed companies would pay for, roughly twenty-six thousand stars on a product a few months old, and the only monetisation attached is a voluntary Founder's Edition supporter tier. Voluntary payment is not pricing power; it is goodwill with a checkout page. Copyleft closes the hosted-service path that usually rescues this position.

The realistic outcomes are a commercial edition, a pivot into the team product this is obviously a wedge for, or an acqui-hire in which the app becomes a hiring credential. Likely acquirer: a terminal or cloud development vendor buying the attention. Position: adopt the app, assume the company becomes something else.

Sixty seats cost nothing, and I still cannot standardise on it, because macOS is the only supported platform and my engineering organisation is not. A tool a third of the team cannot install is a tool I support twice. There is no headless mode, so it contributes nothing to a pipeline and produces no artefact anyone downstream consumes. No directory integration, no audit log, no retention statement.

Onboarding would be trivial for the people who could run it. Not yet. Revisit if a Linux build appears, and I will still need something to show an auditor.

The licence is copyleft with optional commercial terms in the header, and GitHub's detector throwing up its hands at that is annoying but not disqualifying: the source is there and a fork stays open. What I actually want is the control surface, and it delivers, because every action is reachable from the command line and from a Unix socket, so my own scripts can open workspaces and my agent hooks can call cmux notify when they need me.

What I do not get is model freedom. My keys and my own hardware have no role here; this hosts other people's agents and lets me automate the hosting.