by HoangP8
A unified CLI to install and update token-saving plugins — RTK, Caveman, CodeGraph, and Context-Mode — for Claude Code, OpenCode, Codex, and Antigravity. Minimal setup. Any OS.
# Add to your Claude Code skills
git clone https://github.com/HoangP8/toklesstokless is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by HoangP8. A unified CLI to install and update token-saving plugins — RTK, Caveman, CodeGraph, and Context-Mode — for Claude Code, OpenCode, Codex, and Antigravity. Minimal setup. Any OS. It has 52 GitHub stars.
tokless's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/HoangP8/tokless" and add it to your Claude Code skills directory (see the Installation section above).
tokless is primarily written in Go. It is open-source under HoangP8 on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh tokless against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
Save tokens on AI coding agents — no performance loss.
One tool, no config — works the moment it lands.
| Next week (22/6 - 28/6) |
|---|
| Fix Codex & AGY hooks for better MCP (codegraph + context-mode) compliance rateAdd Pi supportAdd Cursor support |
Saving tokens shouldn't be complicated. Many great plugins exist to help coding agents save context and perform better — but setting them up and keeping them updated is hard, especially for non-technical users.
tokless is the lazy solution. One command, pick your agent, restart — IDE and CLI both wired. Cross-platform: macOS, Linux, Windows. No config, no manual edits, re-runnable.
Each agent wired per its own config spec. Pick one, some, or all:
tokless # interactive: pick agents
tokless --agents claude,opencode # wire just these
tokless --agents claude,opencode,codex,antigravity # all
| Tool | What it does | Example |
|---|---|---|
| RTK | Trims noisy bash/tool output | ls -la 45 lines (~800 tok) → rtk ls 12 lines (~150 tok) |
| Caveman | Makes the agent answer in terse prose | 69-tok explanation → 19-tok answer (useMemo) |
| CodeGraph | Query a code graph, skip whole-file reads | 1 codegraph explore call = 0 file reads on 25,874-file repo |
| Context-Mode | Run heavy work in a sandbox, return only what matters | 700KB log → 3KB summary, raw bytes never enter context |
Each tool from official source. Each targets a different waste source — no overlap, no conflict.
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/HoangP8/tokless/main/scripts/install.sh | bash
Windows (PowerShell):
irm https://raw.githubusercontent.com/HoangP8/tokless/main/scripts/install.ps1 | iex
tokless Install + wire everything (default; safe to re-run)
tokless update Show version diff, upgrade the four tools
tokless doctor Show what's wired; warn about broken bits
tokless uninstall Remove everything tokless touched
tokless self-update Update the tokless CLI itself
Flags:
--agents <list> Subset: claude,opencode,codex,antigravity
--dry-run Preview, no writes
--verbose Every step
Restart agents after install so they pick up new config.