Turn Claude, GPT-4, Gemini, and any AI into a full software development company that builds your projects autonomously.
CRAM ESSENTIAL orchestrates multiple AI models as a virtual software development company with 12 specialized roles, 4 orchestration strategies, and a full SDLC pipeline — all inside VS Code.
One click. Multiple AIs. Complete projects.
| Feature | CRAM ESSENTIAL | Single AI Tools |
|---|---|---|
| Multiple AIs working together | Yes | No |
| AI team with specialized roles | 12 roles | None |
| Full project lifecycle (SDLC) | 6 stages | None |
| AI meetings with debates | Yes | No |
| Decision tracking | Yes | No |
| Works with Claude + GPT + Gemini | All at once | One at a time |
Your AI company has specialized employees:
| Role | Default AI | Responsibility |
|---|---|---|
| CEO | Claude | Vision & strategy |
| CTO | Claude | Technology decisions |
| Lead Architect | Claude | System design |
| Senior Developer | GPT-4 | Core implementation |
| Frontend Developer | GPT-4 | UI/UX code |
| Backend Developer | GPT-4 | APIs & databases |
| QA Engineer | Gemini | Testing & bugs |
| Security Auditor | Gemini | Vulnerability analysis |
| DevOps Engineer | Gemini | CI/CD & deployment |
| SEO Specialist | GPT-4 | Search optimization |
| Marketing Strategist | Claude | Go-to-market |
| Performance Engineer | Gemini | Speed optimization |
- Collaborative — Architect designs, Developer codes, Reviewer critiques, then refine. Best for complex projects.
- Sequential — Pipeline chain: each AI builds on the previous output. Best for step-by-step tasks.
- Parallel — All AIs work independently, then results are synthesized. Best for speed.
- Competitive — AIs compete, a judge scores and picks the best. Best for quality.
Discovery --> Architecture --> Implementation --> QA --> SEO/Marketing --> Deployment
Each stage runs an AI meeting where your team:
- Debates approaches and trade-offs
- @mentions each other to ask questions
- Makes tracked decisions with rationale
- Produces code, docs, and reports
Real multi-turn conversations between AI team members. They debate algorithms, review each other's code, challenge assumptions, and reach consensus — just like a real development team.
Add any OpenAI-compatible API: Ollama, Mistral, Together AI, LM Studio, Cohere, and more.
"cramEssential.customProviders": [
{
"name": "ollama",
"baseUrl": "http://localhost:11434/v1",
"model": "llama3",
"apiKey": ""
}
]- Install CRAM ESSENTIAL from the VS Code Marketplace
- Click the CRAM ESSENTIAL icon in the Activity Bar (left sidebar)
- Add at least one API key (Claude, GPT-4, or Gemini)
- Run
CRAM ESSENTIAL: Start Company Projectfrom the Command Palette
You only need one AI provider to get started. Add more for multi-AI collaboration.
| Command | Description |
|---|---|
CRAM ESSENTIAL: Start Company Project |
Run your project through the full SDLC pipeline |
CRAM ESSENTIAL: Start Meeting |
Start an AI team meeting on any topic |
CRAM ESSENTIAL: Set Orchestration Strategy |
Switch between Collaborative/Sequential/Parallel/Competitive |
CRAM ESSENTIAL: Assign Company Roles |
Reassign which AI plays which role |
CRAM ESSENTIAL: View Decision Log |
See all decisions made by your AI team |
CRAM ESSENTIAL: Export Session |
Save the full session to disk |
CRAM ESSENTIAL: Welcome |
Open the welcome & setup guide |
| Setting | Description | Default |
|---|---|---|
cramEssential.claude.apiKey |
Anthropic API key | — |
cramEssential.gpt.apiKey |
OpenAI API key | — |
cramEssential.gemini.apiKey |
Google AI API key | — |
cramEssential.defaultStrategy |
Default orchestration strategy | collaborative |
cramEssential.meetingMaxTurns |
Max conversation turns per meeting | 8 |
cramEssential.maxTokensPerRequest |
Max tokens per AI request | 4096 |
cramEssential.timeoutMs |
Timeout per request (ms) | 120000 |
cramEssential.logLevel |
Logging level | info |
| Provider | Where to get it | Free tier? |
|---|---|---|
| Claude | console.anthropic.com | Yes (limited) |
| GPT-4 | platform.openai.com | No |
| Gemini | aistudio.google.dev | Yes |
MIT - Alen Thomas