rostr

rostr is an open-source CLI from Rosebud Labs that assigns each coding agent session a name from a roster you set, such as 🧙 Gandalf, and places it at the front of the session title. It works with Claude Code, Codex, tmux and herdr, and runs locally or in a cloud mode. The latest GitHub release listed is v1.0.2.

What It Is

rostr is a small CLI plus a few hooks and a skill for your agent. When a session starts, a hook runs rostr draft so the session title shows the agent's name and the agent knows who it is. When the session ends, a hook runs rostr bench to free the name for reuse. The agent is also taught to reach out as that identity, so messages on Slack, GitHub or email can be traced to the session that sent them.

How the Roster Workflow Works

You run rostr assemble (or ask your agent to) to pick a pre-built roster or create your own list of names, with an optional emoji per agent. rostr link gives a specific repo its own roster, and rostr stats shows who is drafted, what they are working on, and who is on the bench. Settings cover emoji display, auto-drafting and how long an unseen draft lasts before release.

Local and Cloud Modes

Local mode is the default, and the site says nothing leaves your machine. Cloud mode is for agents running in the cloud or across machines; the site says no account is needed and that task and repo metadata are end-to-end encrypted. You can enable it during assembly or later with rostr cloud on.

Building on the Output

rostr stats --json and a cloud list API endpoint return rosters, drafted agents and idle agents, which you can use for your own dashboards, monitoring or integrations.