Just Tell the AI What You Want. It Builds the Automation.

What You’ll Find in This Section

• AI Assistant Chat
• Local AI Models
• Cloud AI via OpenRouter
• AI Self-Correction
• AI Vision Page Analysis
• Network API Interception
• MCP Server Integration

The AI Assistant Chat — Your Automation Co-Pilot

Opening the AI Assistant feels like having an automation expert sitting next to you. You type what you want in plain English, and RTILA X translates your goal into a working automation project. The assistant understands context, remembers project details, and asks clarifying questions only when it needs them.

You can describe a multi-step workflow the same way you would explain it to a colleague: “Go to this site, search for these keywords, collect the first 100 job postings, export the results to Google Sheets.” The AI analyzes the page, picks reliable selectors, adds waits where needed, and assembles the full command flow. When it finishes, you review the work and make adjustments in the visual builder if you want.

What Can the AI Assistant Do?

• Generate complete automation projects from natural language
• Analyze web page structure and suggest optimal selectors
• Convert old project formats to the latest version
• Debug and fix broken automations
• Explain what any command or selector does
• Generate data extraction configurations

100% Offline AI — Your Data Never Leaves Your Machine

RTILA X runs local AI models directly on your hardware. That means the words you type in the assistant, the pages you analyze, and the scripts the AI creates all stay on your computer. No data goes to a third-party server, and no hidden cloud processing happens in the background.

Local 7B-8B parameter models typically need 4-6GB of RAM. RTILA X includes integrated management for threads, context size, batch size, and GPU layers so you can tune the model to match your machine.

Supported Local AI Models:

• RTILA Lite 1.5 Assistant 9B, a custom-tuned model included with RTILA X
• Any GGUF model from Hugging Face
• Full control over threads, context size, batch size, and GPU layers
• Models including DeepSeek R1, Qwen 2.5, Llama 3.x, Mistral, and Phi-3

Why Local AI Matters:

• Privacy: no data touches third-party servers
• Cost: zero per-token costs
• Offline: works without internet
• Speed: no network latency

Cloud AI via OpenRouter — When You Want More Power

Local AI fits most privacy-sensitive workflows. When you need a larger model or a specialized reasoning engine, you can connect to OpenRouter directly from RTILA X. The app handles the API connection, model selection, and context management, so you get access to DeepSeek, Claude, Gemini, GPT, and hundreds of other models without building any integration code.

AI Self-Correction — The AI That Fixes Itself

Every automation hits an unexpected page state eventually. When a script fails, RTILA X does not just throw an error and stop. The AI analyzes the error, rewrites the offending code, and retries the operation — up to three times. Enhanced warning logs detail each self-correction attempt so you can see exactly what changed and why.

AI Vision Page Analysis — The AI That Sees

Instead of relying on raw HTML alone, RTILA X can take a screenshot of a page and annotate every interactive element with numbered boxes. The AI “sees” the layout the way a human would, then generates selectors based on that visual understanding. This works especially well on dynamic pages where the underlying DOM structure is messy or changes frequently.

Network API Interception — Skip the DOM Entirely

Many modern sites load data through background API calls. RTILA X passively listens to those XHR and Fetch requests, captures JSON responses, detects object arrays, and generates nested datasets and extraction commands automatically.

You do not need to configure selectors or reverse-engineer endpoints by hand. The intercepted traffic becomes a structured blueprint for your extraction commands, which is faster and more stable than parsing a heavy DOM.

MCP Server Integration — Connect to Any AI Tool

RTILA X includes a Model Context Protocol server. That means compatible tools such as Cursor, OpenAI, and Claude Desktop can discover and control your automations through a standard interface. No glue code and no custom integrations are required.

How It All Works Together

1. You open the AI Assistant and describe the task in plain language.
2. RTILA X analyzes the target page using AI Vision when needed.
3. The assistant generates commands, selectors, and extraction configurations.
4. You review and adjust the automation in the visual builder.
5. During execution, Network API Interception captures structured data where available.
6. If a step fails, AI Self-Correction rewrites it and retries up to three times.
7. The finished automation can run locally, connect through MCP, or be exported as a standalone bot.