KenKaiii /gg-framework

The coding agent that gets more out of your AI subscription. Fastest and leanest of 20 agents tested, with unlimited local search of real open-source code.

★ 303⑂ 56👁 4⑃ 123● 29▤ 351
TypeScriptAGPL-3.0Created 2026-03-04www.skool.com/raid
Download: gg-coder v0.113.1gg-coder-0.113.1 ↓0.113.12026-11-15
+1
Daily
303
stars
56
forks
4
watchers
123
prs
29
issues
351
releases

README

Cause the other agents piss me off.



  English | 简体中文 | 日本語 | 한국어 | Español | Português | Deutsch | Français



  
  
  
  
  
  
  
  


GG Coder is a desktop coding agent that gets more out of the AI subscription you already pay for.

 [!TIP]
 Want the step-by-step guide? The full GG Coder setup, a free 7-day build challenge and 20+ more projects are in my free community: skool.com/raid

 ⭐ If GG Coder saves you time or money, give it a star. It's the main way other developers find it.

 Why GG Coder

- More work per plan. Fastest and leanest of 20 agents I raced on the same model. Claude Code used 169% more for the same work.
- Real code, not guesses. Agent Steroids searches real, current open-source code on your machine. No rate-limited MCP server, no limits.
- Done means checked. Ken Autopilot keeps the work going until it's actually verified.
- Ship it properly, even as a beginner. Coding is only part of it. One-click checks for security, compliance, data safety, accessibility, tests and more, plus a Project health score, so you don't need to know what to ask for.
- Your subscriptions. Claude, ChatGPT, Kimi Code, MiniMax and MiMo plans, any API key, or local models.

| Agent              | Time for the full suite | Usage per suite | Time vs GG Coder | Usage vs GG Coder |
| ------------------ | ----------------------- | --------------- | ---------------- | ----------------- |
| GG Coder       | 102 s               | $0.014      |                  |                   |
| Codewhale          | 112 s                   | $0.022          | 10% slower       | 58% more          |
| Goose              | 132 s                   | $0.020          | 29% slower       | 46% more          |
| GitHub Copilot CLI | 136 s                   | $0.025          | 33% slower       | 83% more          |
| Kimi Code          | 138 s                   | $0.024          | 36% slower       | 74% more          |
| Claude Code        | 142 s                   | $0.037          | 39% slower       | 169% more         |
| ForgeCode          | 155 s                   | $0.042          | 52% slower       | 204% more         |
| Dirac              | 160 s                   | $0.029          | 57% slower       | 114% more         |
| Crush              | 165 s                   | $0.037          | 62% slower       | 173% more         |
| Factory Droid      | 167 s                   | $0.025          | 64% slower       | 83% more          |
| Mistral Vibe       | 169 s                   | $0.036          | 66% slower       | 164% more         |
| DeepSeek-Reasonix  | 170 s                   | $0.027          | 67% slower       | 99% more          |
| Oh My Pi           | 170 s                   | $0.031          | 67% slower       | 123% more         |
| Xum                | 172 s                   | $0.041          | 69% slower       | 203% more         |
| Docker Agent       | 173 s                   | $0.021          | 70% slower       | 51% more          |
| Qwen Code     

Latest release

gg-coder v0.113.1 v0.113.1 2026-11-15
What's new

- Running models on llama.cpp now just works with images. I check whether your model can see before sending a screenshot or picture, so vision models get them and text-only ones never choke on them.

Download the installer for your platform from the assets below.

Snapshot: 2026-11-16T09:23:14Z · Data license Apache-2.0 · JSON