Your coding agents, one cockpit.

You type what you want. An agent does the work — on your hardware, in your repository — and comes back to you before anything you would want to be asked about.

01

Every session, one screen. Your agents appear as live cards — what each one is doing, which one is blocked, one place to aim your words.

02

It adopts what is already running. One small binary on your machine picks up the tmux session you already have open. Nothing restarts, nothing moves, no harness to migrate into.

03

The work happens on your machines. Agents run where your code already lives. Nothing is copied to us to be worked on, and your agent talks to its own model on your own subscription.

04

Any agent, any vendor. Pulsive does not supply the intelligence; it conducts whatever you already use. Switching models later costs you nothing here.

When an agent reaches something a machine should not decide alone — a migration, a deploy, anything destructive — it stops and holds there. Nothing runs until you answer, and it will wait as long as it takes.

asked first

The cockpit goes loud the moment a session needs you — a question, an approval, an error — and quiet again when you have answered.

on the record

Every send and every question answered, timestamped, per session.

step 1

Sign in. Give it your email; a six digit code arrives. No password to choose and none to forget.

step 2

Connect a session. The app walks you through connecting the machine your code is on and gives you a one time code to do it with. About a minute, once per machine.

step 3

Type what you want. Pick the session, write it in plain words, send. It lands exactly as written — no rewriting in between.

step 4

Stay in the loop. When a session wants something that matters it stops and asks, and the cockpit makes sure you see it.

› open pulsive