Schemity is a BYOK ERD tool - bring your own key. Its AI Chat assistant is powered by your own model provider API key, not a shared service baked into the app. This is how Schemity offers AI ERD without cloud lock-in: you choose which provider sees your prompts - or, with Ollama, no provider at all.
AI Chat is a desktop feature. It is not available in the web version, and it is disabled once a trial expires; an active desktop license keeps it on.
Why BYOK
Most AI ERD tools route your schema through their servers. With BYOK:
- You choose the provider and the model.
- The call goes from your machine to that provider, using your key.
- There is no Schemity-operated middle layer holding your data.
That keeps AI-assisted work aligned with the rest of the tool’s promise: private AI database design.
Supported providers
You can add a key for any of these vendors (each has a sensible default model you can change):
- OpenAI (default
gpt-4o) - Claude (Anthropic) (default
claude-sonnet-4) - Grok (xAI) (default
grok-3-mini) - Gemini (Google) (default
gemini-2.5-flash) - DeepSeek (default
deepseek-chat) - Ollama (local models running on your own machine - a base URL instead of an API key)
You can store more than one key - including two for the same vendor - and switch between them.
Adding a key
Open the AI API Keys drawer from the gear icon in the AI Chat panel, then add a key:
- Choose Add API Key.
- Fill in the Name (a label such as
My OpenAI key). - Pick the Vendor.
- Paste the API Key. The field shows the expected format for that vendor, for example
sk-...for OpenAI orsk-ant-...for Claude. - Choose Add.
To change a key, delete it from the list and add it again - there is no inline edit.
Local models with Ollama
Ollama is the one provider that does not need a key. Instead, you point Schemity at your Ollama server’s base URL - the local default works out of the box, and you can change it if your server runs elsewhere. Schemity lists the models installed on that server and streams responses live, exactly like the cloud providers.
With Ollama, AI Chat runs without any data leaving your machine: no API key, no provider account, no cloud. Combined with local JSON storage, that makes Schemity a fully offline AI ERD tool.
Choosing a provider and model
The AI Chat panel has a provider dropdown and a model dropdown in its header. When you pick a provider, Schemity fetches that vendor’s available models and caches the list for about a week (falling back to a built-in list if it cannot reach the provider). Your last-used provider and model are remembered the next time you open AI Chat.
Where your key is stored
Your API key is stored in your operating system’s keychain, not in the diagram files and never in plain text on disk. Only a small amount of provider metadata (the label and vendor) is kept locally. The key leaves your machine only to authenticate directly with the provider you chose.
If you do not configure a key, AI Chat simply stays off - the rest of the offline ERD tool works exactly as before.
Cost control
Because you use your own key, you pay your provider directly at their rates and see your usage in their dashboard. There is no Schemity surcharge for AI, and no token or cost meter inside the app - manage your spend from the provider’s side.
Next
Put it to work in AI-Assisted Design.