Skip to main content 
████████╗ ██████╗  ██████╗ ██╗     ██╗  ██╗██╗████████╗      ██████╗██╗     ██╗
╚══██╔══╝██╔═══██╗██╔═══██╗██║     ██║ ██╔╝██║╚══██╔══╝     ██╔════╝██║     ██║
   ██║   ██║   ██║██║   ██║██║     █████╔╝ ██║   ██║  █████╗██║     ██║     ██║
   ██║   ██║   ██║██║   ██║██║     ██╔═██╗ ██║   ██║  ╚════╝██║     ██║     ██║
   ██║   ╚██████╔╝╚██████╔╝███████╗██║  ██╗██║   ██║        ╚██████╗███████╗██║
   ╚═╝    ╚═════╝  ╚═════╝ ╚══════╝╚═╝  ╚═╝╚═╝   ╚═╝         ╚═════╝╚══════╝╚═╝

> Where LLMs Collaborate, Not Compete

The universal orchestration layer that makes Claude, Gemini, Codex, and Qwen work together. Stop fighting with your AI. Start conducting it.

🎃
Launching in:
00d 00h 00m 00s

October 31, 2025 • Midnight PDT (Halloween)

No inference tax • We don't scrape your code • Full context, no BS

The Spec-Context Revolution

toolkit-cli is different. Before any command, it reads your project's DNA—the specs, plans, and contracts that define your vision. It doesn't just fix code; it preserves your intent.

BEFORE: Generic AI 

# Error: 'user_name' can be None
# Generic AI "fix":
user_name = get_user_name() or ""
# Just silences the error ❌

Standard AI tools ignore your specs and apply generic fixes that break your business logic.

AFTER: toolkit-cli 

# Spec: "Invalid users must re-register"
# toolkit-cli fix:
user_name = get_user_name()
if user_name is None:
    return redirect_to_registration()
# Preserves your intent ✓

Reads your specs first. Fixes code while preserving business logic and design decisions.

Every command. Every agent. Spec-aware.

A Conductor's Toolkit

35 slash commands. Each one spec-aware. Each one orchestrating multiple LLMs. Each one designed to make you unstoppable.

⚡
10s

/oneshot

Idea to production-ready codebase in 10 seconds. Backend, frontend, database, CI/CD—fully scaffolded.

🔐
TUI

/keys

Beautiful terminal UI for managing .env files. Auto-masks secrets so you never expose them by accident.

💩
QA

/bs

Pre-commit quality gate. Catches architectural drift, mock code, and scope creep before they hit main.

⚡
AUTO

/implement

Execute all tasks from tasks.md automatically. Sit back while your backlog clears itself.

🔧
SPEC

/fix

Spec-aware code fixes. Reads your requirements first, then fixes bugs without breaking your business logic.

🎯
AI

/next

Smart recommendations for what to work on next. Analyzes errors, git status, tests, and specs.

🎓
NEW

/tutorial

Interactive onboarding with hands-on examples. Learn YOLO mode, multi-agent, and the complete workflow in minutes.

⏪
NEW

/undo

Time-travel for your codebase. Auto-checkpoint before risky operations, rollback changes with one command.

👥
MULTI

/peer-review

3+ AI agents review your code in parallel. Get architectural, performance, and security insights simultaneously.

+24 more commands: /specify /plan /debug /improve /tutorial /undo /license
View Full Command List

One more thing.

We said 35 commands. We weren't kidding.

🔍
/analyze
Cross-artifact validation
💩
/bs
Detect violations
💬
/clarify
Refine specs
📜
/constitution
Define principles
🔍
/debug
Root cause analysis
🚨
/errors
Screenshot detection
🔧
/fix
Spec-aware fixes
💡
/help-me
AI pair programmer
⚡
/implement
Execute tasks.md
🔨
/improve
Multi-agent review
🚀
/init
Initialize project
🔐
/keys
.env manager
📚
/learn
Knowledge base
⚒️
/make
AI implementation
🔍
/new-feature
Discover features
🎯
/next
Smart recommendations
⚡
/oneshot
Idea to production
👥
/peer-review
Multi-agent review
📋
/plan
Implementation roadmap
🪞
/reflect
Code critique
🔒
/security
Threat modeling
📝
/specify
Feature specs
✅
/tasks
Generate tasks.md
🧪
/test
Testing strategy
🎨
/ux
UX design
ℹ️
/version
System status
🎓
/tutorial
Interactive onboarding
⏪
/undo
Rollback changes
📜
/license
License management
📊
/usage
Usage tracking
🎨
/mock-ups
Visual design
🧹
/clean-up
Code cleanup
⚠️
/weaknesses
Threat modeling
✨
/polish
Code polishing
🔄
/re-context
Rebuild context

35 commands. Zero configuration. Infinite possibilities.

Patent Pending

Get Started →

Works with Every AI

Use toolkit-cli with Claude Code, Cursor, Windsurf, Roo, or any editor you love. One tool. 12 AI agents. Infinite combinations.

~/my-project
$ toolkit-cli fix auth.py --ai "claude gemini qwen codex"
● Claude analyzing architecture patterns...
● Gemini reviewing UX implications...
● Qwen checking i18n compliance...
● Codex generating optimized solution...
✓ Synthesized fix applied. 4 agents collaborated.
Claude
Claude
Sonnet, Opus, Haiku
Gemini
Gemini
Pro, Flash, Ultra
💧
Qwen
2.5, Coder, Math
OpenAI
OpenAI
GPT-4, o1, Codex
Copilot
Copilot
GitHub AI
▶️
Cursor
AI-first IDE
🌊
Windsurf
Codeium IDE
🦘
Roo
Roo Cline
Ollama
Ollama
Local models
🔀
OpenRouter
Unified API
🤖
Auggie
AI Assistant
💻
Kilocode
Code AI
🔓
OpenCode
Open source AI
💻

Claude Code

Native slash commands

# In Claude Code chat
/fix auth.py
⚡

Any Terminal

Direct CLI access

$ toolkit-cli fix auth.py
🚀

VS Code, Cursor...

Task integration

# tasks.json
"command": "toolkit-cli"
Provider agnostic. Zero lock-in. Pure freedom.

Mix and match models. Switch providers. Use local or cloud. Your workflow, your rules.

The 3 Wise Men

Claude + Codex + Gemini = AGI

--ai claude gemini codex

I use Claude, Codex, and Gemini for everything. Not because one is better—but because three minds are smarter than one.

Claude sees structure. Codex sees patterns. Gemini sees scale. Together, they catch what any single model misses.

This isn't about benchmarks or leaderboards. It's about collaboration. When three agents debate, synthesize, and build consensus—that's when you get something closer to AGI than any single model alone.

"The wisest human doesn't work alone. Why should your AI?"

Pay Once. Use Forever.

No subscriptions. No recurring charges. Multi-agent collaboration that pays for itself.

Individual Enterprise

FREE

$0 One-time
  • check_circle 30-day unlimited trial
  • check_circle 2 hours/day forever after
  • check_circle Community Support
MOST POPULAR

DEVELOPER

$20 One-time
  • check_circle 5 hours/day usage
  • check_circle Unlimited agents
  • check_circle Unlimited projects
  • check_circle Community Support

PROFESSIONAL

$100 One-time
  • check_circle 10 hours/day usage
  • check_circle Unlimited agents
  • check_circle Unlimited projects
  • check_circle Community Support

UNLIMITED

$200 One-time
  • check_circle Unlimited usage
  • check_circle Unlimited agents
  • check_circle Unlimited projects
  • check_circle Community Support

All tiers include lifetime access to updates. Usage limits reset daily. Try free for 30 days—unlimited.

How We Got Here

Not a software engineer. For 30 years, tried to learn to code—dyslexic hands and ADHD kept my fingers two words behind my mind. Became a software architect instead. Built startups. Hired 300+ developers on oDesk/Upwork over 2 decades. 80,000 billed hours. Learned from PRs and screenshots, watching developers from CERN, NASA, and Microsoft build what these hands couldn't.

2016, tried to invent agentic AI. Lost millions. Lost friends. Lost everything. Last startup, Findy—Perplexity's search intelligence combined with Scale AI's data labeling, powered by public typeahead transformers—years before Perplexity existed. Lost everything again.

August 2024, started over. Claude Dev became Cline, then Cursor, then Roo. Forked Kilo, built an AI OS, made a fucking mess. But 25 billion tokens later, figured out the gaps in context. Testing Amazon's Kiro—saw the potential, but the mistakes led to spec-kit. Toolkit was born.

Every line of code written by AI. Built in Claude Code in the terminal—no IDE required. Alpha, but already production-capable.

49 now. Living at mom and dad's. Licensing toolkit to build a snowboard factory in the mountains. Make boards with friends. Live the dream that's been 30 years in the making.

If this helps you build your dreams, you just helped me live mine.

Never fucking quit.

Thanks to Tom Latzo, Fredrick Karlsson, Serge Gulin, and Vladimir Glafirov for teaching me what I couldn't learn on my own.

Aaron Rosenthal
aka roseyballs

check_circle Copied to clipboard!