AI Assistant

inSCADA offers two solutions that connect AI assistants directly to your SCADA system. Both solutions use the same 39 tools to read live values, monitor alarms, write scripts, analyze historical data, generate charts, and more.

Note: Both solutions require inSCADA JDK11 edition.

Choose Your Solution

	MCP Server	AI Assistant (Desktop)
Platform	Claude Desktop, VS Code Copilot, Cursor	Standalone Windows application
Setup	.mcpb or JSON config	Installer (.exe)
AI Provider	Claude (subscription)	Claude, Ollama, Gemini, Groq, etc.
Data privacy	Data passes through Claude system	100% local with Ollama
Cost	Claude Pro/Max subscription	API token cost (free with Ollama)
Interface	Within Claude Desktop	Built-in chat UI (TR/EN)
Details	MCP Server	Sections below

Capabilities

• Natural language querying — Query your SCADA data by speaking in Turkish or English
• Live data & charts — Create line, bar, gauge, and forecast charts
• Alarm analysis — Query active alarms and analyze alarm history
• Script development — Write, test, and deploy Nashorn ES5 compatible scripts
• Excel export — Export query results as .xlsx files
• API discovery — Discover and call 625+ inSCADA endpoints in natural language

39 Tools — 8 Categories

Category	Count	Description
Space & Data	10	Space, project, variable, variable search, script, connection management
Animation	2	Animation listing and details
SCADA Operations	10	Live values, alarms, connection status, historical data
Charts	5	Line, bar, gauge, multi-series, forecast charts
Custom Menu	6	Menu CRUD operations (template-based creation supported)
Generic API	3	Discover and call 625+ endpoints
Export	1	Excel file generation
Guide	2	Script rules, animation element details, best practices

AI Assistant — Desktop Application

Standalone Windows application. Choose your own AI provider, run fully local with Ollama — your data never leaves your computer.

Supported AI Providers

Provider	Description
Claude (Anthropic)	Recommended. Ideal for complex SCADA analysis and long script generation
Ollama (Local)	Open-source models running on your computer. No internet required, data stays local
OpenAI Compatible	Google Gemini, Groq, Cerebras, OpenRouter, and other OpenAI-compatible API providers

Installation

1. Download the inSCADA AI Assistant installer from inscada.com/download
2. Run the installer and complete the setup
3. Launch the application

Requirements: Windows 10/11, a running inSCADA server (JDK11 edition)

Initial Configuration

Configure the application from the settings page on first launch:

1. AI Provider Selection:

• Claude: Enter your Anthropic API key
• Ollama: Enter Ollama server address (default: http://localhost:11434)
• OpenAI Compatible: Select provider, enter API URL and key

2. inSCADA Connection:

• inSCADA URL: Server address (e.g. http://localhost:8081)
• Username: inSCADA login username
• Password: inSCADA login password
• Test Connection: Use the test button to verify connection settings

Features

• Chat interface — Full Turkish and English support, conversation history
• Excel export — Download query results as .xlsx files
• Live gauge — Auto-refreshing gauge panels every 2 seconds
• Forecast charts — AI predictions visualized on the same chart with historical data
• Script development — Write, test, and deploy scripts — all from a single chat
• Confirmation mechanism — Approval dialog for dangerous operations

Security

Dangerous Tools

The following tools affect real equipment and require user confirmation on every use:

• inscada_set_value — Writes values to real equipment (PLC, inverter, etc.)
• inscada_run_script — Executes server-side scripts
• update_script — Modifies script code
• inscada_api (POST/PUT/DELETE) — Modifies data via generic API

In MCP Server: These tools are marked with MCP safety annotations (destructiveHint: true). The AI client prompts for user confirmation before calling them.

In AI Assistant: Blocked server-side and queued in pendingActions. The user approves or cancels via the confirmation dialog in the UI.

Data Privacy

• Credentials are stored only on the local machine
• AI Assistant binds to 127.0.0.1 only — no network access
• When using Ollama, no data leaves your computer
• In MCP Server, session tokens are kept in memory and discarded on exit

Usage Examples

A few example prompts:

Live value:

claude space, Energy Monitoring Demo, ActivePower_kW value?

Chart:

claude space, Energy Monitoring Demo, ActivePower_kW last 24 hours chart

Script writing (guide required):

inscada guide read. claude space, Energy Monitoring Demo,
write a script that sends notification when temperature exceeds 60°C every 10 seconds

Excel export:

claude space, Energy Monitoring Demo, export ActivePower_kW 24 hour data to Excel

For detailed scenarios, prompt templates, and token optimization: Usage Guide