An AI assistant that actually sees Inventor.
No separate ChatGPT tab to copy-paste from – a native chat panel that reads your model, calls tools, and checks its own work.
BYO-AI: your model, your choice
27 models across 14 providers directly selectable – Claude, GPT, Gemini and more – or run fully local and free via Ollama/LM Studio. You decide which AI sees your CAD data. See all models →
Upload images & sketches
Attachment button or Ctrl+V: a photo or screenshot lands directly in the chat. The foundation for a sketch-to-part workflow – Kidge reads dimensions and asks when something is unclear.
Self-verification, not claims
For multi-step tasks, Kidge first plans a task list with success criteria, works through it step by step, and backs up every step with concrete evidence. "Done" is only accepted once everything is proven.
Session management for long chats
Automatic history compaction past a token threshold (/verlauf), real token usage display (/tokens), and an adjustable auto-continue budget (/limit) keep even long sessions manageable.
Persistent memory
Kidge remembers lessons across sessions – mistakes avoided, successful approaches, API knowledge – and pulls in tool and guide search when needed, instead of re-guessing every edge case.
Web search for current questions
Via /help and /search, Kidge can also research beyond the Inventor API – for example, general usage questions about Inventor itself.