Zentrun — One Prompt, Complete Automation

"Send yesterday's sales to Slack every morning at 9 AM" This single sentence becomes working automation.

No setup wizards. No workflow builders. No API configurations. Just describe what you want.

🎯 What You Can Build

Business Intelligence

"Collect daily sales data, compare to targets, identify trends, send executive summary"
→ Complete BI pipeline with data collection, analysis, and reporting

Data Analysis & ML

"Analyze customer churn patterns and build prediction models"
→ Data exploration + ML training + predictive insights

Competitive Monitoring

"Track competitor websites for pricing changes, analyze impact, suggest responses"
→ Web monitoring + analysis + strategic recommendations

Customer Intelligence

"Monitor support tickets, classify by urgency and topic, predict escalation needs"
→ Ticket analysis + ML classification + predictive insights

Content Operations

"Find trending topics, generate content ideas, schedule posts, track performance"
→ Research + content creation + scheduling + analytics

Everything that you want can be automated.

🎬 See It Working

Real demo: Complete marketing automation from prompts

Watch a marketing workflow get built with just natural language:

1. Collects AI news from multiple sources
2. Summarizes content with AI analysis
3. Posts automatically to Twitter
4. Analyzes engagement patterns
5. Visualizes results in dashboard

Each step generated from language. No manual configuration.

🚀 Installation

# Or build from source
git clone https://github.com/andrewsky-labs/zentrun
cd zentrun
yarn && yarn dev

💭 Why Automation Still Feels Like Work

The "AI-Assisted" Reality

Zapier with AI

• ❌ "AI helps you build workflows" → Still need to configure triggers, actions, connections
• ❌ "Natural language setup" → AI generates workflow, you still need to understand their interface

Make.com with AI Agents

• ❌ "AI Agents automate tasks" → Agents work within workflows you build with drag-and-drop
• ❌ "Conversational automation" → Conversation helps configure, doesn't replace configuration

Power Automate with Copilot

• ❌ "Describe your workflow" → Generates Power Platform flows, still need to set up connections
• ❌ "Natural language flows" → You still debug in their visual interface

Zentrun's Approach

💬 "Monitor competitor prices and alert me when they change significantly"
→ Running automation. No setup. No configuration. No workflow building.

💬 "Analyze customer feedback, identify trends, create weekly summary"
→ Complete pipeline from data collection to analysis to reporting.

💬 "Track our mentions on social media, classify sentiment, escalate negatives"
→ Full monitoring and response system with ML classification.

The difference: Others help you build automation. Zentrun just automates.

🔧 How This Is Possible

Browser-Native Execution

• Uses your existing browser sessions (no API keys needed)
• Inherits all your login credentials automatically
• Handles complex web interactions like a human
• Bypasses API limitations and rate limits

AI-Powered Code Generation

• Converts natural language to executable Python/SQL/JavaScript
• Generates browser automation scripts
• Creates data analysis workflows
• Builds ML training and inference pipelines

Local Processing Power

• Fast local database (SQLite/DuckDB) for data operations
• Built-in ML libraries for custom model training
• Real-time analysis without cloud dependencies
• Complete workflow execution on your machine

Intelligent Adaptation

• Self-healing when websites change
• Learning from successful automation patterns
• Context-aware decision making
• Automatic error recovery and retry logic

🛠️ Technical Architecture

Natural Language Processing

• Advanced prompt parsing and intent recognition
• Context-aware code generation for automation logic
• Multi-step workflow planning from single descriptions
• Error recovery and script refinement

Browser Integration

• Chrome DevTools Protocol for seamless browser control
• Session inheritance for instant access to logged-in services
• JavaScript execution environment for dynamic interactions
• Anti-detection techniques for reliable web automation

Data & ML Pipeline

• SQLite/DuckDB for high-performance local analytics
• Vector databases for embedding storage and retrieval
• scikit-learn, TensorFlow, PyTorch for ML workflows
• Real-time inference and model serving capabilities

Execution Engine

• Parallel processing for complex multi-step automations
• Intelligent scheduling and resource management
• Fault tolerance with automatic retry mechanisms
• Version control and rollback for automation logic

📋 Getting Started

Start Simple

1. Install Zentrun
2. Try: "Summarize my unread emails from today"
3. Watch it work in your browser
4. Add more complex automations gradually

Common Use Cases

• Daily email and calendar summaries
• Social media monitoring and analysis
• Competitive intelligence gathering
• Customer feedback analysis
• Financial reporting automation

Advanced Workflows

• Multi-source data integration pipelines
• Custom ML model training on your data
• Complex decision-making automations
• Real-time monitoring and alerting systems
• Reusable MCP workflows (Save as Zent)

🤖 Supported Model Providers

Ollama	Deepseek	Silicon	QwenLM
Doubao	MiniMax	Fireworks	PPIO
OpenAI	Gemini	GitHub Models	Moonshot
OpenRouter	Azure OpenAI	Qiniu	Grok

** Compatible with any model provider in OpenAI/Gemini/Anthropic API format

✨ System Requirements

Minimum specs for a decent experience:

macOS: 13.6+ (8GB RAM for 3B models, 16GB for 7B, 32GB for 13B) Windows: 10+ with GPU support for NVIDIA/AMD/Intel Arc Linux: Most distributions work, GPU acceleration available

✨ Contributing

Please refer to Contribution Guide.

📃 License

LICENSE

Stop building workflows. Start describing what you want. Automation that actually understands natural language.

⭐ Star if you want automation to work like you always thought it should!