An AI copilot, built into Inventor.
Kidge is a native chat panel for Autodesk Inventor. Describe what you need in plain language – Kidge reads your models, sets features, manages assemblies and drawings, and verifies its own work instead of just guessing.
Bring your own AI.
Kidge doesn't lock you into an AI subscription. 34 models across 16 providers are already directly selectable – or run fully local and free via Ollama. You decide which AI sees your data and what it costs.
34 models, 16 providers
Claude, GPT, Gemini and more – freely selectable in the settings dialog, switch anytime.
Fully local & free
Via Ollama or LM Studio, Kidge also runs entirely without the cloud and without ongoing costs.
Connect Claude Code & Codex as backend
Already have a Claude Code or Codex console running? You can now connect it directly as Kidge's AI backend – no separate API key needed, just your existing subscription, with real model choice inside each CLI. A bit slower than direct API access, but at no extra cost.
Not a chatbot on the side – an assistant that actually works inside Inventor
Kidge runs as a native add-in inside the Inventor process itself: full API access, real exceptions instead of caught dialogs, no fragile scripting from the outside.
Native integration
Runs directly inside the Inventor process instead of external COM automation – more stable, faster, with real API access.
Broad tool coverage
79 tools from iProperties through sketches, features and material to assemblies, drawings, sheet metal, iLogic, and Vault – one consistent toolset.
Choose your own model (BYO-AI)
Claude, GPT, Gemini, free cloud models, or fully local via Ollama – you decide what fits your data and your budget.
Self-verification
For multi-step tasks, Kidge plans first, works step by step, and backs up every step with evidence – instead of just claiming success.
From download to your first task in minutes
Install
Download and run the installer – Kidge registers itself automatically as an Inventor add-in.
Choose a model
Add an API key, or go completely free and local with Ollama – no cloud required.
Open the chat
The "Kidge" button in the ribbon opens the chat panel right next to your model.
Describe the task
Say what needs doing in plain language – Kidge plans, executes, and checks its own work.
More than just Inventor
Kidge is starting focused as a CAD AI for Autodesk Inventor. Expanding to other CAD systems is on the long-term roadmap.
Frequently asked questions
What is a CAD AI?
A CAD AI is an AI assistant that works directly inside a CAD program – comparable to ChatGPT, but with real access to models, sketches, features and drawings instead of just text. Kidge is a CAD AI for Autodesk Inventor.
Is there an AI copilot like ChatGPT built directly into Autodesk Inventor?
Yes – that's exactly what Kidge is: a native chat panel inside Inventor that takes tasks in plain language and carries them out with 79 dedicated tools directly against the Inventor API, instead of just generating text.
Does Kidge also work with SolidWorks, FreeCAD or other CAD systems?
Kidge is currently focused on Autodesk Inventor. Expanding to SolidWorks, FreeCAD, Autodesk Fusion 360 and Solid Edge is on the long-term roadmap, but not available yet. Autodesk Vault already has a first, generic connection today (listing and running Vault commands from chat) – deeper, native integration is planned for later.
Which AI models does Kidge support?
Kidge isn't locked to a single provider: 34 models across 16 providers are currently selectable, from Anthropic (Claude), OpenAI (GPT) and Google (Gemini) to Mistral, DeepSeek, xAI, Azure OpenAI, OpenRouter, and locally via Ollama or LM Studio – including free and fully local options.
What does BYO-AI mean for Kidge?
BYO-AI stands for "bring your own AI": Kidge doesn't lock you into a specific AI provider or a subscription. Freely choose from 34 models across 16 providers, run fully local and free via Ollama or LM Studio – and you can even connect an already-running Claude Code or Codex console directly as your AI backend, with no separate API key needed.