crouter
Using crtr

Your first crtr session

When you have installed crtr and want to start an agent conversation, read this page because it shows the one command that creates a root node and opens its viewer.

Your first crtr session

You can start an agent in a terminal, talk with it in a tmux pane, then close that pane without stopping the agent. Use a root node for a conversation that you will drive yourself.

Install and set up

Install crouter globally, then run the setup program once on the machine.

npm install -g @north-light/crouter
crtr sys setup

crtr sys setup opens an installer. It signs you in to a model provider, then offers companion packages and system dependencies. It also refreshes crouter's tmux bindings.

Start a conversation

From the directory where you want the agent to work, create a root node with a clear request.

crtr node new --root "Inspect this repository and report the failing tests."

--root creates an independent node rather than a child of another agent. It starts the node's detached broker. When you run the command inside tmux, it also opens a viewer pane for that node; type your next message there to continue the conversation. Outside tmux, the command prints the node id instead. Open the viewer in your terminal with:

crtr surface attach to <node-id>

The viewer is not the agent process. The broker keeps the node's conversation and session running separately from the pane, so closing the terminal or leaving the viewer does not discard the work. A root node's final report is not sent back to the terminal that created it; open the node again when you want to see its result.

flowchart LR
  terminal[Your terminal] --> command[crtr node new --root]
  command --> broker[Node broker]
  broker <--> viewer[tmux viewer]
  broker --> canvas[Canvas state]

Return to a node

The create command prints the node id. Keep it if you expect to return later. From tmux, focus that node again with:

crtr surface node focus <node-id>

For a running node, this brings its viewer into view. For a parked resident node, it opens the saved conversation; sending a message wakes it. For finished work, it opens a read-only saved transcript instead of starting another agent process.

For the controls inside that viewer, continue with The viewer. For a view of every node, use Canvas and dashboard.