MCP Workbench

🚀 Unified Interface for LLM Providers & MCP Servers

✨ Features • 🚀 Quick Start • 📷 Screenshots • 🌐 Providers • 📊 Performance • 🗺️ Roadmap

---

📖 Overview

MCP Workbench is a production-ready platform for working with multiple LLM providers, MCP servers, and data science tools. Built with Next.js, React, and TypeScript.

Feature	Details
Providers	13 providers (OpenAI, Anthropic, Google, Ollama, HuggingFace, Replicate, etc.)
MCP Registry	100+ servers from GitHub with one-click install
Performance	85%+ cache hit rate, 80% faster APIs with Redis
Security	160+ blocked patterns, sandboxed execution
Accessibility	WCAG 2.1 AA compliant with keyboard navigation

---

✨ Features

💬 Multi-Provider Chat 13+ LLM providers with seamless switching Persistent history & tool integration Syntax highlighting & Markdown rendering Model selection with one-click 📚 MCP Registry & Tools Browse 100+ MCP servers One-click installation with auto-runtime detection Visual server management interface Configuration UI for all servers 🐍 Python Notebook Auto-detect environments (uv, conda, venv) Rich output with images & plots Integrated terminal for commands Artifact export capabilities

🤖 Model Browser Browse all models from connected providers Filter by capability (vision 👁️, embeddings 🔮, image-gen 🎨) Reasoning detection & metadata Quick "Use in Chat" integration ⚡ Performance & Caching Redis caching with 85%+ hit rate 80% faster API responses Database indexes & query optimization Real-time performance monitoring 🏥 Health Dashboard Live system metrics (DB, memory, disk) Auto-refresh with countdown Cache statistics & hit rates Resource usage tracking

---

🌐 Supported Providers

Local (3)

Provider	Description
🦙 Ollama	Open-source LLM runtime
🎬 LM Studio	Local model server
⚙️ Custom	Your own endpoint

Remote (10)

Provider	Models	Capabilities
🤖 OpenAI	GPT-4, GPT-3.5, DALL-E	Vision, Chat, Image-Gen, Audio
🧠 Anthropic	Claude 3.5 Sonnet/Opus	Vision, Chat, Reasoning
🔮 Google AI	Gemini 2.5 Flash/Pro	Vision, Chat, Reasoning
⚡ Groq	Mixtral, Llama	Ultra-fast inference
🌐 OpenRouter	100+ models	Access all models
🤗 HuggingFace	1000+ models	Community models, Embeddings
🔄 Replicate	Llama, SDXL, FLUX	Image-Gen, Audio, Vision
🤝 Together AI	50+ models	Open-source models
🧪 Mistral AI	Mistral models	Fast inference
💬 Cohere	Command models	Chat, Embeddings

🎯 Specialized Capabilities

• 👁️ Vision: GPT-4V, Claude 3, Gemini, LLaVA
• 🔮 Embeddings: text-embedding-3, nomic-embed, all-MiniLM
• 🎨 Image-Gen: DALL-E, SDXL, Stable Diffusion, FLUX
• 🎤 Audio: Whisper v1/v2/v3, wav2vec2
• 🧠 Reasoning: GPT-4, o1-preview, Claude 3, Gemini Pro

---

🚀 Quick Start

Installation

# Clone repository
git clone https://github.com/scorpiontaj/mcp-workbench.git
cd mcp-workbench

# Install dependencies
bun install

# Configure environment
cp .env.example .env.local
# Edit .env.local with your API keys (optional - can set via UI)

# Initialize database
bun db:push
bun db:seed

# Start development server
bun dev

Open http://localhost:3000

Configuration

Option 1: UI (Recommended)

1. Go to Settings → Providers
2. Click any provider card
3. Enter API key & test connection

Option 2: Environment Variables

# Add to .env.local
OPENAI_API_KEY="sk-..."
ANTHROPIC_API_KEY="sk-ant-..."
GOOGLE_API_KEY="AI..."
# ... more providers

Get API Keys

• OpenAI - GPT-4, GPT-3.5
• Anthropic - Claude
• Google AI - Gemini
• HuggingFace - Community models
• Replicate - Llama, SDXL

---

📷 Screenshots

Dashboard - Real-time Monitoring	Chat - Multi-Provider Interface
Models - Browse & Filter	Registry - 100+ MCP Servers
Notebook - Python Environment	Health - System Metrics

View all screenshots →

---

📊 Performance

Optimization Results (v4.0)

Metric	Before	After	Improvement
Cache Hit Rate	0%	85%+	Fixed ✅
API Response	~500ms	~100ms	80% faster ⚡
Component Renders	100%	50%	50% reduction 📉
Time to Interactive	2.5s	1.5s	40% faster 🚀

Lighthouse Scores

• Performance: 95+ 🟢
• Accessibility: 98+ 🟢
• Best Practices: 100 🟢
• SEO: 100 🟢

---

🛠️ Tech Stack

Frontend: Next.js 16, React 19, TypeScript 5.9, Tailwind CSS 4.1
UI: shadcn/ui (Radix UI), Lucide Icons
Backend: Next.js API Routes, PostgreSQL, Drizzle ORM
Caching: Redis (ioredis)
State: Zustand, SWR
Performance: React.memo, code splitting, virtualization

---

🔑 Why MCP Workbench?

✅ Universal Access - 13 providers, one interface
✅ Multi-Modal - Vision, audio, embeddings, image-gen
✅ Blazing Fast - 85%+ cache hit rate, Redis-backed
✅ Secure - 160+ blocked patterns, sandboxed execution
✅ Beautiful UI - Premium design, mobile-responsive
✅ Python-First - Auto-detect environments
✅ Production Ready - Error boundaries, health monitoring
✅ Accessible - WCAG 2.1 compliant
✅ Developer Friendly - Integrated terminal, command palette

---

🗺️ Roadmap

🎯 Phase 3: Chat Enhancements (In Progress)

Feature	Priority	Status
Chat export (PDF, Markdown)	⭐⭐⭐⭐	📋 Planned
Message reactions	⭐⭐⭐	📋 Planned
Message search within chats	⭐⭐⭐⭐	📋 Planned
Chat templates	⭐⭐⭐⭐	📋 Planned
Conversation branching	⭐⭐⭐	📋 Planned

📊 Phase 4: Monitoring & Analytics

Feature	Priority	Impact
Model performance metrics	⭐⭐⭐⭐⭐	Track response time, token usage, costs
Response quality tracking	⭐⭐⭐⭐	Monitor success/failure rates
Cost analysis dashboard	⭐⭐⭐⭐	Compare API costs across providers

🚀 Phase 5: Advanced Features

Feature	Priority	Description
Side-by-side model comparison	⭐⭐⭐⭐	Compare outputs from multiple models
Batch testing	⭐⭐⭐⭐	Send same prompt to multiple models
Performance benchmarks	⭐⭐⭐⭐	Compare response quality and speed

🔌 Phase 6: Integration & API

Feature	Priority	Status
Webhook support	⭐⭐⭐	External system integrations
GraphQL endpoint	⭐⭐⭐	More flexible API queries

---

🎯 Phase 2 Features (Active Development)

✨ Chat Templates System ✅

Pre-built AI personas for common tasks:

• Code Review Assistant - Detailed code feedback and analysis
• Data Analysis Expert - Dataset analysis and pattern detection
• Creative Writing Coach - Storytelling and narrative assistance
• Research Paper Assistant - Academic research support

🎯 Message Reactions & Annotations ✅

Provide feedback and annotate conversations:

• 6 Reaction Types: Helpful, Unhelpful, Bookmark, Love, Insightful, Excellent
• Annotation Types: Notes, Highlights (with colors), Flags, Important markers
• Features: Track reactions, edit annotations, organize by color

🔍 Advanced Search in Chats ✅

Find messages across all conversations:

• Full-Text Search with relevance ranking
• Smart Filters: Role, Provider, Token count, Date range
• Auto-Complete with real-time suggestions
• Search History & Popular search terms
• Analytics: Track search behavior and trends

📊 Upcoming Phase 2 Features

• Model Comparison Interface (in development)
• Analytics Dashboard (planned)
• Vector Database & RAG (planned)
• Team Collaboration (planned)

Status: 3/9 Phase 2 features complete. See PHASE2_PROGRESS.md for details.

📝 Progress Summary

Phase 1: ✅ 100% Complete Phase 2: 🚀 50% Complete (3/6 features)

• ✅ Chat Templates
• ✅ Message Reactions & Annotations
• ✅ Advanced Search
• ⏳ Model Comparison
• ⏳ Analytics & Usage Tracking
• ⏳ Vector Database & RAG

Total: 45+ features completed, 6+ in active development

Note: Feature requests welcome! Check Issues or Discussions to suggest features.

---

🤝 Contributing

Contributions welcome! Fork the repo, create a feature branch, and open a PR.

git checkout -b feature/amazing-feature
git commit -m 'Add amazing feature'
git push origin feature/amazing-feature

---

📄 License

MIT License - see LICENSE file for details.

---

📞 Support

• Issues: Report bugs or request features
• Discussions: Join the community
• Portfolio: scorpiontaj.me

---

⭐ Star this repo if you find it helpful!

Made with ❤️ by Tajeddine Bourhim

📦 View on GitHub