babo

Run plain-language `.babo` scripts compiled by Claude Code

TLDR

Run a Babo script (builds on first use, uses cache afterward)

$ babo hello.babo

Force a rebuild after editing the source description

$ babo build hello.babo

Check whether the cached build is still current

$ babo check hello.babo

Show cache metadata for a script

$ babo info hello.babo

Remove all cached builds

$ babo clean

SYNOPSIS

babo [run] file.babo [args...]babo check|build|info|clean file.babo

babo is a Python CLI that turns `.babo` files — short natural-language program descriptions — into runnable Python programs. On first run it sends the description to Claude Code (`claude -p`), which generates an entry point, optional `requirements.txt`, and an isolated virtual environment under `.baboc/<name>.baboc/` next to the source file (similar in spirit to Python's `_pycache_`).Subsequent runs skip the build when the `.babo` file is older than its cache directory. Arguments after the script path are forwarded to the generated program. A shebang `#!/usr/bin/env babo` lets `.babo` files execute directly once babo is on PATH.Babo scripts can call other `.babo` modules via a generated `runtime.call_babo()` helper, each module keeping its own venv so dependencies do not collide.

PARAMETERS

run

Explicit run (default when the first argument is a `.babo` file).

build

Force regeneration from the `.babo` source, ignoring a fresh cache.

check

Print whether the cache is FRESH or STALE.

info

Show build metadata (paths, timestamps, package list).

clean

Delete every `.baboc/` cache directory under the current tree.

CONFIGURATION

No global config file. Each `.babo` file owns a `.baboc/<stem>.baboc/` cache with `venv/`, `requirements.txt`, `metadata.json`, and the generated `babo` executable entry point.

CAVEATS

Requires Python 3.10+, the claude CLI installed and authenticated, and network access during builds. Builds are not offline. The tool is experimental and depends on Claude Code API availability.