neetorecord setup <assistant> writes the rules file that teaches an AI coding
assistant how to use the CLI: how authentication works, what the global flags
do, how output is shaped, and how to discover commands it has not seen before.
Set up an assistant
Every subcommand except
setup claude writes into the current directory, so
run it from the root of the project you want the assistant to use NeetoRecord
in. setup cursor and setup windsurf overwrite their rule file when it
already exists. setup copilot, setup gemini, and setup codex keep the
rest of the file and replace only the section between the markers, so there is
never a duplicate. Run the command again after every CLI upgrade so the rules
match the installed commands.
Claude Code
setup claude needs Claude Code to have been run at least once, so that
~/.claude/ exists. It replaces ~/.config/neetorecord/claude-plugin with a
fresh copy of the plugin, which holds the NeetoRecord skill, a session start
hook that reports whether the CLI is signed in, and a /neetorecord-doctor
command. It then prints the slash commands to run inside Claude Code:
What makes the CLI work well for agents
Self describing
neetorecord commands prints the whole command tree with every flag as
JSON. An assistant can read it instead of guessing at a flag.Fewer tokens
--toon carries the same data as the JSON envelope in a compact encoding,
which keeps large listings from filling the context window.Predictable errors
Every command exits non-zero on failure and writes one line to stderr, so a
failure is easy to detect and to act on.
Chainable output
--quiet prints just the identifier after a create, which is what an
assistant needs to feed the next command.