Effortless: the free Claude Code plugin that picks the effort for every prompt

Your quick questions cost the same as building a whole app. Effortless is a free, open-source Claude Code plugin that has Haiku 5.5 judge each prompt and set the reasoning effort, so quick questions stop costing as much as big builds. This is the full version of the SAVE guide from my Reel. Facts checked against the README on 20 Oct 2026.

01What it is

Effortless is an MIT-licensed Claude Code plugin by HeyCubit. Before each prompt runs, a judge (Haiku 5.5 by default) reads it and picks the reasoning effort: Low for quick questions, High for hard jobs. A bar above the prompt shows the effort it picked, how full the chat is, and how long the prompt cache stays warm, with one-click Compact and Handoff when a chat gets heavy.

02Why it matters

Most people set one effort level and leave it there, so a one-line question gets the same deep thinking as a full refactor. Every extra step re-reads the whole chat, and that re-read is most of the cost. Lower effort on easy prompts means fewer steps.

In the author's own benchmark (4 coding tasks, 3 runs each, 96 runs on 10 Oct 2026, through Anthropic's API), Opus 5.5 with Effortless cost $0.149 per task against $0.190 at medium, $0.209 at high and $0.257 at xhigh: 21%, 29% and up to 42% less, with all 12 answers correct.

03Setup

Type these in Claude Code's chat box

/plugin marketplace add HeyCubit/effortless
/plugin install effortless@effortless
/reload-plugins

Run the two /plugin commands above, then /reload-plugins (or restart Claude Code).
A short setup opens above the prompt. Keep Haiku as the judge: no key needed, it runs on your own Claude login.
Run /effortless setup to redo it, or /effortless settings to change anything later.

Prefer a terminal?

claude plugin marketplace add HeyCubit/effortless; claude plugin install effortless@effortless

04Good to know

The savings are small-sample: four mostly easy tasks, three runs each. On hard work it picks more effort and saves less. The runs were fresh API chats, not long chats on a plan.
Effort only changes on Opus 5.5 and Sonnet 5.5; on older models Auto pauses. On Sonnet the author measured even vs medium, 10% saved vs high and 35% vs xhigh.
Changing effort yourself turns Auto off at your next message. You always win.
"Cheaper model when it can" (moving simple prompts to Haiku) is off by default: switching models mid-chat rewrites the prompt cache and usually costs more.