Cause the other coding agents piss me off.
The only coding agent you can walk away from.
macOS · Windows · runs on the AI plan you already pay for
I built this because every other coding agent pissed me off.
They all do the same thing: you sit there and babysit. Approve this. Confirm that. Read the output, spot the mistake, tell it again. You're not building, you're supervising. And the second you close the laptop, everything stops.
So I built the one I wanted. Two things nobody else has.
Hook it to a Telegram chat and your laptop becomes something you text. Send a message, send a voice note, get told when it's done. From the pub, from bed, from a queue at the shops.
Voice notes get transcribed on your own machine, not sent to anyone. Put a job on a timer and it checks your site every 15 minutes and fixes whatever broke while you were asleep.
Every other agent marks its own homework. This one doesn't.
Flip on Autopilot and Ken, a whole second agent, reviews every finished job. Not good enough? He hands it straight back with exactly what's wrong, and it goes again. And again. Until it's right.
That's the real thing, mid-loop. Ken caught a bug that would have quietly let through four times the traffic limit, sent it back, and signed off on the fix. Nobody typed a single thing in between.
Signed and notarized on macOS. It updates itself, so you install it once and forget about it.
Chat stays readable. Every file it touches and every command it runs streams in a panel at the bottom, so nothing happens behind your back and your conversation never turns into a wall of noise.
Run as many as you like at once, each on its own project and its own model, all reviewed by Ken:
Your agent learned to code from a snapshot of the internet, and that snapshot is old. So before GG Coder writes anything nontrivial, it reads real, current open-source repos sitting on your own disk, via Agent Steroids. Offline, no rate limits.
One click on the Home screen installs it. Then /steroids profiles your project, finds the repos that match it, and indexes the ones you pick. Your agent stops guessing at APIs that changed last quarter.
You know the wall: mid-build, and your usage limit hits. Everything stops for five hours.
GG Coder can hold a subscription and a backup key at the same time. The plan goes first, and the key takes over the second it runs dry. It just keeps going. A live meter up top shows exactly how much you've burned and when it resets, so it's never a surprise.
And nothing is locked in: swap models mid-conversation, or run one on your own machine with no internet at all.
| It can see | Drag in a screenshot, paste a design, throw a video at it. It watches the video and builds what you showed it |
| Plan mode | It looks around and writes you a plan. Nothing gets touched until you say go |
| Catches its own mistakes | Broken code is spotted and fixed in the same breath it was written, before it ever reaches you |
| Knows your CLIs | Spots 35 platform tools like gh, vercel and railway in your project and drives them for logs, deploys and env vars instead of sending you to a dashboard |
| Picks up where you left off | Finds the projects you've been working on in Claude Code and Codex too, not just GG Coder ones |
| Remembers your project | Notes, memory, chat export, and your own shortcut commands. Add any tool you find online by pasting one line |
| Watches your usage | A live meter up top shows how much of your plan you've burned and exactly when it resets. No surprise cut-offs |
| A bit stupid, on purpose | XP, ranks and streaks for shipping. Sound. ASCII banners. Coding should be fun |
Prefer the terminal? Same agent, same engine:
npm i -g @kenkaiiii/ggcoder
ggcoderOAuth login so there are no API keys to paste, full terminal UI, tools, MCP, LSP diagnostics, session resume. → packages/ggcoder
The desktop app forks zero agent logic. Windows, IPC and UI live in gg-app/; everything else is the exact same spine the CLI runs, and every layer ships on npm on its own.
GG Coder desktop app ⭐
└── @kenkaiiii/ggcoder (CLI + app sidecar)
├── @kenkaiiii/gg-ai (standalone)
├── @kenkaiiii/gg-agent ──► @kenkaiiii/gg-ai
└── @kenkaiiii/gg-core ──► @kenkaiiii/gg-ai
| Package | What it does |
|---|---|
@kenkaiiii/gg-ai |
One streaming API for every provider up there |
@kenkaiiii/gg-agent |
Agent loop with multi-turn tool execution |
@kenkaiiii/gg-core |
Shared guts: model registry, OAuth, auth storage, paths |
@kenkaiiii/ggcoder |
The CLI, plus the sidecar the desktop app runs |
👨💻 Run it from source
git clone https://github.com/KenKaiii/gg-framework.git
cd gg-framework
pnpm install
pnpm --filter @kenkaiiii/ggcoder build # build the sidecar first
cd gg-app && pnpm tauri devpnpm build # build all packages (gg-ai → gg-agent + gg-core → ggcoder)
pnpm check # typecheck
pnpm test # vitest
pnpm lintTypeScript 5.9 · pnpm workspaces · Tauri 2 · React 19 · Vite 7 · Ink 6 · Vitest 4 · Zod v4
Packaging (bundled Node runtime, single-file sidecar, code signing) is in
gg-app/DISTRIBUTION.md. README art is generated by
node gg-app/scripts/render-readme-art.mjs; product shots by
node gg-app/scripts/capture-screenshots.mjs.
- YouTube @kenkaidoesai, tutorials and demos
- Skool community
MIT licensed. Use it, change it, ship it.
Every model. Every project. One window each.
