Getting started

Tillerly runs its agents on your team's machines. The dashboard is where your team plans, starts and signs off the work, and each member pairs a Runner, the process that runs the agents on their own machine.

What each member needs

• A Mac or a Linux machine. Windows is not supported, and on Linux or WSL2 the sandbox needs bubblewrap.
• Node 22.12 or later.
• git, logged in to your code host over SSH, for the projects Tillerly clones for you.
• Claude Code, the CLI claude or the VS Code extension, logged in with your Claude account. An Anthropic API key, stored on the harness page of the settings, works too.

Before the agents can work

The dashboard shows a checklist with this title above the settings, until every step is done.

1. Pair your Runner. Start it on your machine and pair it from the dashboard. The page on the Runner says how.
2. Choose the harness and its model. Tillerly checks, through your Runner, that the harness answers and has a default model.
3. Set the folder of your repositories. The folder where your clones are, or will be.
4. Enable a project and choose where its tickets are. A project is one repository, and its tickets come from a Tickets connector, Jira today. The page on connectors lists them.

Once the four are done, your team can start a ticket in the Builder, design a change in Plans, or let Triage read your production errors.