tokless: Make AI Coding Agents More Efficient and Effective in 30 Seconds

Phan Hoang — Sun, 05 Jul 2026 07:51:04 +0000

Many great tools for coding agents already exist.

The painful part is discovering, installing, updating, and combining them without conflicts.

tokless is the lazy CLI that connects everything with one command.

tokless integrates six popular packages (kaparthy skills, caveman, ponytail, rtk, codegraph, context-mode) for improving AI coding agents in under 30 seconds.

Choose your agents, install, restart, and start coding.

https://github.com/HoangP8/tokless

Six tools, one setup — automatically configured without conflicts
Ready in under 30 seconds — select, install, and restart
Zero manual configuration — no configuration files to edit by hand
Cross-platform — macOS, Linux, and Windows
One-command updates — tokless update upgrades everything together

Installation

macOS and 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

After installation, start the interactive setup:

tokless

Or select agents directly:

tokless --agents claude,opencode

To configure every currently supported agent:

tokless --agents claude,opencode,codex,antigravity

Supported AI coding agents

Agent	Status
Claude Code	✅ Supported
OpenCode	✅ Supported
Codex	✅ Supported
Antigravity	✅ Supported
Pi	In progress
Cursor	In progress
Factory Droid CLI	In progress
GitHub Copilot	In progress

Integrated tools

Tool	Purpose
karpathy-skills	Encourages simple solutions, minimal changes, validated assumptions, and goal-driven execution
caveman	Produces shorter responses while preserving important technical information
ponytail	Favors minimal code, standard libraries, and no speculative features
rtk	Compresses terminal output before it enters the model's context
codegraph	Retrieves code relationships from a local code graph instead of repeatedly reading entire files
context-mode	Processes large tool outputs separately and returns only the information the agent needs

Each tool solves a different problem, zero conflicts.

Restart your coding agents after installation so they reload their configuration.

Issues and compatibility reports: https://github.com/HoangP8/tokless/issues

If tokless is useful to you, pls star it.

tokless: minimal setup for token and context optimization in long agent sessions

Phan Hoang — Fri, 19 Jun 2026 09:24:54 +0000

Hello everyone, I built this package https://github.com/HoangP8/tokless with one simple philosophy: "saving tokens shouldn't be complicated. While there are many great plugins out there to help coding agents save context and improve performance, setting them up and updating them is so complex—especially for non-technical users."

Therefore, tokless is the ultimate "lazy" solution. Run a single command, click on agents you work with, then all ready for you, no need to configure anything more. Restart your agent to start saving tokens and context window. Fully open-source and minimal setup!

This package will enable 4 tools:

CodeGraph (Code Discovery)
RTK (Terminal Output Filtering)
Context-Mode (Task Execution Sandbox)
Caveman (Minimal Agent Responses)

If you encounter any issues or compatibility problems with your specific setup, pls raise an issue. If you find the work useful to you, I hope you can give it a star!