tokenolo

The tokenolo CLI for the tokenolo token-optimization service: send a prompt, get back a shorter, token-cheaper version that preserves intent and every constraint.

tokenolo paraphrase "Could you please, if it's not too much trouble, summarize this for me?"
#   STDOUT: Summarize this.
#   STDERR: saved 9 tokens, -60% (15 → 6, mode: balanced)

This is the client. tokenolo talks to a tokenolo backend (the server holds the upstream LLM key and does the rewrite). Point it at your server with TOKENOLO_API_URL (default http://localhost:8787). Running the backend is covered in the repo.

Install

npm i -g tokenolo      # installs the `tokenolo` (and `tk`) command

Requires Node ≥ 20.

Commands

# Authenticate. Run bare to be prompted (hidden), or pipe the key via stdin
# (keeps it out of argv / shell history). The key is stored 0600 and never printed.
tokenolo login
echo "$TOKENOLO_KEY" | tokenolo login

# Who am I + remaining daily quota.
tokenolo whoami

# Paraphrase. Optimized prompt → stdout, savings summary → stderr (pipes cleanly).
tokenolo paraphrase "…"
echo "…" | tokenolo paraphrase                 # read prompt from stdin
tokenolo paraphrase --mode aggressive "…"      # aggressive | balanced (default) | safe
tokenolo paraphrase --json "…"                 # full JSON response on stdout

# Forget the stored credential.
tokenolo logout

--url <url> overrides the backend for any command; otherwise the URL resolves as --url → TOKENOLO_API_URL → the URL stored at login → http://localhost:8787. (tk still works as a short alias for every command.)

Security

The CLI only ever holds your scoped tk_live_… key (stored 0600 at ~/.config/tokenolo/credentials.json). The upstream LLM key lives only on the server and is never shipped to the client. Keys are hashed at rest, rate-limited, daily-token-capped, and instantly revocable.

Links

• Source & backend: https://github.com/merkhanov/tokenolo
• Issues: https://github.com/merkhanov/tokenolo/issues

License

MIT