Terminal commands

Create independent tasks, inspect state, capture output and send input through the CLI. For a one-shot build or test, run the actual command as the task so its exit status represents that work.

Create

anytty terminal create --endpoint local --json \
  --name docs-check --cwd /absolute/project/path \
  --tag project=demo --env CI=1 \
  -- sh -lc 'printf "check started\n"; pwd'

Replace the directory with a real target path. This example requires sh. Arguments after -- select the program. Omit them for the default shell; --attach attaches after creation, and --rows / --cols set initial dimensions.

Retain the response’s target for subsequent commands.

Inspect and wait

anytty terminal show ENDPOINT:TERMINAL_ID --json
anytty terminal wait ENDPOINT:TERMINAL_ID --state exited --timeout 2m --json
anytty terminal capture ENDPOINT:TERMINAL_ID --lines 200 --json

wait observes lifecycle state, not log-based readiness. After reaching exited, inspect exit_code. Timeout does not terminate the task.

Capture reads projected history by default; --live reads the latest native screen. --lines and --cols affect formatting, not the running PTY size.

Send input

Only after confirming an idle shell:

anytty terminal send ENDPOINT:TERMINAL_ID --literal 'pwd' --enter

send delivers input to the current program; it is not an independent remote execution API. It supports text, named keys, hex or stdin according to mutually exclusive input rules. --key Ctrl-C can interrupt a program and is not a connectivity probe.

Lifecycle actions

restart begins a new run, kill ends a task and remove deletes a record. Consult subcommand help for parameters. Collect status and required output before cleanup. See scripting and results.