Tickets, red builds, review fixes, production errors, Slack requests, dependency upgrades. Helmline hands them to your coding agents across every project, around the clock, on your machine. You step in only where it takes a person, and one screen tells you what happened.
Here's one night of it, in the real dashboard on made-up sample data. The clock follows your scroll.
Label an issue agent, or let Plan the day read your backlog and propose what to arm. Nothing starts that you didn't hand over.
It works the issue and opens a pull request. It never merges, never pushes to a protected branch, never touches production.
A new error lands for Globex. Its lead wakes on its own, on Globex's own account, reproduces the timeout and starts on a fix.
Initech's lead keeps exiting. Helmline sees the loop, says so in plain words, and lets it rest for 40 minutes instead of burning your budget at 2am.
A pull request fails CI. The agent reads the log, fixes the stale test and pushes. Green by 02:46.
Four pull requests, a fixed production error, about $17 of model spend under the daily cap you set. Initech ran on a local model, for nothing. Your day ribbon fills itself in.
Umbrella's lead needs a decision about refunds. It asks once, proposes a default, and waits under Needs you.
What shipped, what broke and got fixed, what waits on you, what it cost. One screen, told as a story, and the same brief on your phone before you open the laptop.
Switch a ticket on and an agent takes it tonight. Tickets too big for one pull request come back as a proposed split for you to accept.
Lead questions, review requests and Slack in one list, most urgent first. A request becomes a ticket after a veto window you can cancel; replies arrive with a suggested answer.
Acme on your Claude subscription, Globex on Globex's own Bedrock, Initech on a local open model, Umbrella on Codex. Costs land where they belong, with CI, deploys and errors per client beside them.
What it's doing right now, what it found, what it asked. Stop it, or answer it, from the same drawer.
Each account picks its agent CLI and its model: your Claude subscription, a client's own cloud, or a model running on this machine. Clients map to accounts, so their work runs where it should.
Helmline's own design in light or dark, a mind where leads are thoughts, an operation with units in the field, or Tokenmaxx, where they're Ralph loops and pausing is touching grass. Words, characters and maps change with it.
Pick an agent CLI and a model per account, so each client runs on its own key, its own cloud, or a model on your own machine. A local model costs nothing and never touches your budget.
Run end to end through HelmlineWired to its documented command line, not yet run by us
Planned for v1.1, not in the download yet.
Free and source-available. One command installs it on your machine, with your agents, your models and your repositories.
Install with one command
curl -fsSL https://get.helmline.workers.dev | sh
Then open 127.0.0.1:4317. Setup finds your repositories, agent and model, and shows one screen to confirm.