tokenolo
Documentation

Introduction

What tokenolo is and why it exists.

tokenolo is a local proxy for coding agents. It wraps your agent behind a proxy on 127.0.0.1, removes redundant filler from your own messages before they're sent, and forwards everything else to the model untouched. Removal-only, cache-safe, and private — your code never leaves your machine.

Why it exists

Most tools that trim an agent's token usage have to read and rewrite everything it sends — your files, tool output, and history — usually on their own servers. That saves more tokens, but it means routing your code through a third party on every request.

tokenolo takes the opposite bet. It does deliberately less: it trims only the filler from your own messages, forwards the rest verbatim, and keeps your code on your machine. It is a middleware layer, not a new model and not a new workflow.

What you get

  • Private by design — the proxy is local; requests go to the model under your own key, and nothing is stored.
  • Cache-safe — byte-stable, removal-only edits keep your prompt cache valid.
  • A modest token trim — filler removed before send; a small reduction, not a dramatic one.
  • Zero rewiring — drop-in for Claude Code and Codex. Same agent, same commands.

Start here

HumansMachine