Work with your agents from the terminal

What changed

The nebula-ai CLI has a new set of commands for the work you do in a workspace. From your shell you can now:

start a Task for an agent with tasks create, then follow, rename, stop or archive it
send, read, edit and react to messages in a Channel or a Task with messages
create, run and pause scheduled jobs with jobs
list, create and delete miniapps with miniapps
save and remove workspace or personal skills with skills

channels and agents gained the matching management commands too. The CLI reads and writes the same synced workspace as the desktop app, so a Task you start in the terminal appears in the Channel straight away, and your team sees it like any other.

How it works

Install the CLI with npx nebula-ai (or brew install agent-labs-dev/nebula/nebula-ai on macOS and Linux), then run nebula-ai login.
Hand work to an agent on a Channel's roster: nebula-ai tasks create --channel releases --agent quill "Write customer release notes for 2.14 from Diego's update". The CLI prints the new Task's id.
The Task shows up in the Channel and the agent gets to work. Read its replies with nebula-ai messages list --task <id>, or open it in the app.
Post back to the team with nebula-ai messages send --channel releases "...".

Every command takes --json, so you can use them in scripts. The CLI runs on macOS, Linux and Windows.

Why it matters

A lot of work starts in the terminal: a deploy finishes, a script fails, a release is cut. Now you can hand the next step to an agent from the same place, without switching windows, and the result still lands in the Channel where your team works.

The Nebula desktop app's releases Channel beside a terminal. In the terminal, Maya ran nebula-ai tasks create to hand release notes to the agent Quill, then nebula-ai messages send. In the Channel, Quill's finished Task card shows the release notes file, followed by Maya's message.

Download Nebula