Run one command (no REPL)

Use borescope from scripts: run a single command and exit, or pipe a sequence of commands in on stdin.

Most of the time borescope is an interactive prompt. For scripts, aliases, and CI, you can run it non-interactively instead.

Run one command

Pass --command to run a single command and exit, without ever entering the REPL:

$ borescope myapp/0 --command "services"
$ borescope myapp/0 --command "plan"
$ borescope myapp/0 --command "logs -n 100 myapp"

The command is parsed and run exactly as if you'd typed it at the prompt, including pipes:

$ borescope myapp/0 --command "logs -n 500 myapp | grep -i error"

Output goes to stdout, errors to stderr.

Exit codes

borescope's process exit code reflects what happened:

That makes --command safe to use in shell conditionals:

$ if borescope myapp/0 --command "health" >/dev/null; then echo healthy; fi

Pipe a script on stdin

When borescope's stdin isn't a terminal and you haven't passed --command, it reads commands from stdin one line at a time, runs each, and exits with the status of the last:

$ borescope myapp/0 <<'EOF'
services
plan
logs -n 50 myapp
EOF

Or pipe a file:

$ cat checks.borescope | borescope myapp/0

Blank lines are skipped. This is handy for canned diagnostic runbooks you keep in version control.

Running container tools

exec works the same non-interactively, so you can reach any tool in the container from a script:

$ borescope myapp/0 --command "exec cat /proc/1/cmdline"
$ borescope myapp/0 --command "exec ps aux"

Scripting tips

• Quote the whole command. --command "logs -n 100 myapp" is one argument; borescope parses it the same way the REPL does.
• Pick the model explicitly in automation with --model, rather than relying on whatever model happens to be current.
• Prefer --snapshot when you want structured, machine-readable state rather than the human-formatted output of individual commands. See Capture a state snapshot.
• One unit per invocation. borescope targets a single unit; loop in your shell to sweep several.