effortless plugin
A Claude Code plugin that picks the model's effort level for every prompt and shows the prompt cache countdown
- 286 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- A Claude Code plugin runs code on your machine through hooks and starts the `claude` and `git` programs
- A judge model reads the text of each of your prompts, and a TypeSafe key pasted by hand sits in plain text in `~/.config/jev/.env`
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/effortlessDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- A Claude Code plugin runs code on your machine through hooks and starts the `claude` and `git` programs
- A judge model reads the text of each of your prompts, and a TypeSafe key pasted by hand sits in plain text in `~/.config/jev/.env`
This is third-party code. Review the repository files before installing.
What it does
effortless is a Claude Code plugin that frees you from switching the effort level by hand. Before each prompt a judge model reads it in about a second and picks a level: quick questions get Low, hard jobs get High. The default judge is Haiku 5.5 on your own Claude login, and no API key is needed. Short follow-ups such as "go", "ok" or "yes" keep the level already picked without asking the judge. A bar appears above the prompt field. It shows the picked level and the chat's model, how full the context is in percent and a countdown until the prompt cache goes cold. After that the next message pays full price to re-read the whole chat, and the bar warns about it. The Auto button turns the judge on and off, and changing the level by hand in the app turns Auto off until your next message. When a chat gets heavy, Compact and Handoff appear. Handoff writes a chat summary and lets you continue in a clean chat: the quick version takes a few seconds, the full one checks git and saves `HANDOFF.md` or runs your own skill. Every compaction, including `/compact` and the automatic one, is written by Haiku 5.5 by default, and if it fails Claude Code compacts as usual. The bar warns about 5-hour and weekly limits after 80% and offers a Save mode capped at Medium. You can additionally connect TypeSafe's Jev judge with a key, which answers in 0.25 seconds instead of one. Switching to a cheaper model exists in the plugin but is off by default.
Who it is for. For people who work a lot in Claude Code on Opus or Sonnet, watch token and limit use and do not want to pick the effort level by hand each time.
Good fit when
- Some requests to Claude are simple and some hard, and you do not want to switch effort by hand
- You want to see how many minutes remain before the prompt cache goes cold so you continue the chat in time
- A chat has grown and you want to compact it or move to a clean chat with a summary in `HANDOFF.md`
- The 5-hour or weekly limits are near the edge and you need a mode with an effort ceiling of Medium
- You want chat compaction done by the fast, inexpensive Haiku 5.5 model
Not a fit when
- You work outside Claude Code: the plugin lives in its interface and does not work in other agents
- You need a fixed effort level for every request: you would have to turn Auto off
- You need a judge without network calls: scoring goes through Haiku 5.5 or the TypeSafe service when a key is set
- You want to swap in a cheaper model mid-chat: each model keeps its own prompt cache, so switching in a long chat usually costs more
Example request
Install the effortless plugin for Claude Code and run `/effortless setup` to keep Haiku as the judge without a key.Limitations
- The plugin works only in Claude Code and runs code on your machine through hooks. The programs it starts are `claude` (update and uninstall on your button), `git` (new-version check and full Handoff) and the judge itself.
- A TypeSafe key for the Jev judge is optional. A key entered through `/plugin configure` is kept in Claude Code's secure storage, while one pasted in the setup guide is written in plain text to `~/.config/jev/.env`.
- The savings figures were measured by the author on four real tasks with three runs each, 96 runs through the API in all. They are mostly easy tasks; on hard ones the plugin picks more effort and saves less. On Sonnet the result against Medium came out even.
- Switching to a cheaper model is experimental and off. It is enabled in Settings, under Model.
- The repository's second part, textless, is a separate desktop app and is not covered by this card.
How to disable. Run `claude plugin uninstall effortless@effortless`, then `/reload-plugins`. If you entered a TypeSafe key in the setup guide, delete it from `~/.config/jev/.env`. To switch off only the judge use the Auto button on the bar.
FAQ
How do I install effortless in Claude Code?
Run `/plugin marketplace add HeyCubit/effortless` and `/plugin install effortless@effortless`, then `/reload-plugins`. A short setup opens above the prompt field and can be repeated with `/effortless setup`.
How much does effortless save?
In the author's test on four tasks (96 API runs, 10 October 2026) Opus 5.5 cost fell 21% against Medium, 29% against High and 42% against xhigh, with every answer still right. The author notes that the tasks were mostly easy.
Does the judge need an API key?
No. Haiku runs on your Claude login. A TypeSafe key is needed only for the extra Jev judge, which answers in about 0.25 seconds.
What does the bar above the prompt show?
The picked level and model, context fullness, the prompt cache countdown and the reason for the choice. The Auto, Compact and Handoff buttons control the judge and chat compaction.
Why is switching to a cheaper model off?
Each model keeps its own prompt cache, so moving a prompt mid-chat forces the whole chat to be written again. In an active chat such a switch usually costs more, not less.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code plugin: a status line with context, 5-hour and 7-day limits, cost and usage of Codex, Gemini and z.ai
A marketplace of 17 Claude Code mods: token-spend breakdown, a rollback edit timeline, assumption tracking, a verification board and a live command feed
A Claude Code plugin that catches your corrections in chat and, after review, writes them into CLAUDE.md and AGENTS.md
A statusline for Claude Code: token and cost usage, context left, rate limits and git branch in four styles and six themes
pordee (พอดี)
pordee
A Claude Code plugin that makes the agent answer in Thai briefly: no polite particles, hedging or greetings
A marketplace of fourteen plugins for Claude Code power users: model council, email, calendar, video digests and routine tasks
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| effortless (this entry) | no | 286 | |
| claude-dashboard | no | 584 | a status line with context, limits and cost, while effortless also picks the effort level itself |
| Tokentap | no | 811 | a local proxy with a token usage dashboard, with no control over effort level |
README in short
The page presents two Claude Code mods from one author and describes effortless in detail: install, bar elements, judges, keys and trust, benchmark results and why model switching is off. The textless block is about a separate desktop app.