A general-purpose AI agent capable of completing complex, long-horizon tasks. Powered by a multi-agent system with specialized agents for code, research, presentations, and multimedia processing.
CUA is a sophisticated multi-agent system that:
- Decomposes complex tasks into manageable sub-tasks
- Routes work to specialized expert agents
- Coordinates execution with dependency management
- Aggregates results into cohesive outputs
| Agent | Capabilities |
|---|---|
| Code Agent | Full-stack web apps, Auth, Database, Stripe, E2E Testing |
| Research Agent | Web search, Deep analysis, Browser automation, Charts |
| PPT Agent | Beautiful presentations, Flexible layouts, PPTX export |
| Multimodal Agent | Image/Audio/Video input, Generation, OCR |
✅ Full-Stack Development
- Next.js, React, Vue, APIs
- Server Actions, tRPC, REST
✅ Authentication
- NextAuth.js, Clerk, Auth0
- OAuth, JWT, RBAC
✅ Database
- Prisma, PostgreSQL, MongoDB
- Migrations, Relations
✅ Payments
- Stripe Checkout
- Subscriptions, Webhooks
✅ Testing
- Playwright E2E
- Jest, React Testing Library
✅ Comprehensive Research
- Multi-source search
- API integration
- Browser automation
✅ In-Depth Analysis
- Data analysis with code
- Chart generation
- Report compilation
✅ Aesthetics
- Flexible layouts (not just templates)
- Professional design
- Data visualizations
✅ Export Quality
- High-fidelity PPTX
- HTML to PowerPoint
- Animations support
✅ Input
- Long-text files
- Video, Audio, Images
- OCR and document processing
✅ Output
- Image generation
- Audio synthesis (TTS)
- Video analysis
- GitHub/GitLab - Repository management, issues, PRs
- Slack - Messaging and notifications
- Google Maps - Places search, directions
- Figma - Design file access, components
Create any custom MCP from scratch or by wrapping existing tools:
import { createCustomMCP } from "@/lib/mcp";
createCustomMCP(
"my-api",
"My Custom API",
"Access my custom service",
[
{
name: "my_tool",
description: "Does something useful",
inputSchema: { /* JSON Schema */ }
}
]
);┌─────────────────────────────────────────────────────────┐
│ User Request │
└─────────────────────┬───────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ Orchestrator │
│ • Task Analysis • Decomposition • Coordination │
└─────────────────────┬───────────────────────────────────┘
│
┌─────────────┼─────────────┬─────────────┐
▼ ▼ ▼ ▼
┌───────────┐ ┌───────────┐ ┌───────────┐ ┌───────────┐
│ Code │ │ Research │ │ PPT │ │Multimodal │
│ Agent │ │ Agent │ │ Agent │ │ Agent │
└─────┬─────┘ └─────┬─────┘ └─────┬─────┘ └─────┬─────┘
│ │ │ │
▼ ▼ ▼ ▼
┌─────────────────────────────────────────────────────────┐
│ Tools & MCPs │
│ • Code Execution • Web Search • Image Gen │
│ • File Operations • Browser Use • Audio/Video │
│ • Testing • Charts • PPTX Export │
└─────────────────────────────────────────────────────────┘
- Node.js 18+
- API Keys (all FREE tiers):
- Groq - Fast inference
- Google AI Studio - Gemini
- E2B - Code sandbox (optional)
# Clone
git clone https://github.com/sheikhcoders/Computer-Using-Agent.git
cd Computer-Using-Agent
# Install
npm install
# Configure
cp .env.example .env.local
# Add your API keys to .env.local
# Run
npm run dev- / - Chat interface (Lightning/Pro modes)
- /agent - Multi-agent task execution
src/
├── app/
│ ├── api/
│ │ ├── chat/route.ts # Chat API (Lightning/Pro)
│ │ └── agent/route.ts # Multi-agent API (SSE)
│ ├── agent/page.tsx # Multi-agent UI
│ └── page.tsx # Chat UI
├── components/
│ ├── agent/ # Agent UI components
│ ├── chat/ # Chat components
│ └── ui/ # shadcn/ui
├── hooks/
│ └── use-agent.ts # Agent state management
└── lib/
├── agents/
│ ├── types.ts # Type definitions
│ ├── config.ts # Agent configurations
│ ├── orchestrator.ts # Task orchestration
│ ├── code-agent.ts # Code specialist
│ ├── research-agent.ts # Research specialist
│ ├── ppt-agent.ts # Presentation specialist
│ └── multimodal-agent.ts# Multimodal specialist
├── mcp/
│ └── index.ts # MCP integration
├── e2b/
│ └── sandbox.ts # Code execution
└── ai/
├── registry.ts # AI provider registry
├── modes.ts # Lightning/Pro modes
└── prompts.ts # System prompts
# AI Models (Required - Both have FREE tiers)
GROQ_API_KEY=gsk_... # Groq Console
GOOGLE_GENERATIVE_AI_API_KEY=... # Google AI Studio
# Code Execution (Optional - FREE tier)
E2B_API_KEY=e2b_... # E2B Dashboard
# MCPs (Optional)
GITHUB_TOKEN=ghp_... # GitHub API
SLACK_BOT_TOKEN=xoxb-... # Slack API| Category | Technology |
|---|---|
| Framework | Next.js 15 (App Router) |
| AI SDK | Vercel AI SDK v4 |
| Models | Groq (Llama, Gemma), Google (Gemini) |
| Code Execution | E2B Sandboxed Runtime |
| UI | shadcn/ui, Radix UI, Tailwind CSS |
| Streaming | Server-Sent Events (SSE) |
| Icons | Lucide React (SVG only) |
Following strict WCAG 2.1 guidelines:
- ✅ Full keyboard navigation (WAI-ARIA APG)
- ✅ Visible focus indicators
- ✅ Minimum hit targets (24px desktop, 44px mobile)
- ✅
prefers-reduced-motionsupport - ✅ Proper ARIA labels and roles
- ✅ Skip to content link
- ✅ Color contrast (APCA compliant)
- Vision capabilities (screenshot analysis)
- E2B Desktop sandbox (full browser control)
- Voice input (Whisper integration)
- Multi-agent collaboration (parallel execution)
- Task memory and replay
- Custom agent creation
- Follow accessibility guidelines (MUST/SHOULD/NEVER)
- Use SVG icons only (never emoji in UI)
- Test keyboard navigation
- Ensure
prefers-reduced-motionsupport - Write tests for new features
MIT License - see LICENSE
Built by @sheikhcoders
Note: This project uses free API tiers. Please respect rate limits and usage policies.