cmux

You orchestrate multi-pane terminals for AI agents by using a CLI to spawn split panes, send commands, and read screen outputs. This enables automated multi-pane workflows and parallel task execution across different surfaces.

Automating parallel tasks in a terminal workspace is best achieved by spawning split panes for sub-agent tasks. You can send commands to individual surfaces, read outputs, and manage workspaces within a single CLI to coordinate complex dev workflows.

Yes, you can manage browser surfaces and terminal panes from a single CLI. The system supports browser automation and multi-surface coordination, allowing AI agents to interact with both terminal and browser environments simultaneously.

You read terminal output programmatically by sending commands to a surface and capturing the screen output. The system supports JSON-based programmatic output and uses canonical references like pane:NN and surface:NN for environment-aware CLI interactions.

No dependencies are required to use cmux for workspace automation. It operates independently to orchestrate terminals, surfaces, and workspaces, applying environment-aware CLI interactions to manage multi-pane workflows for AI agents.

You should not use a multi-pane terminal orchestrator when your AI workflows require only single-threaded, sequential command execution. It is designed for complex, parallel sub-agent tasks and multi-surface coordination rather than simple linear scripts.