Start diagramming
Open a template or create a diagram, write Mermaid source, then preview it instantly.
Browse templates →Help for your local workspace.
Everything you need to start writing, previewing and exporting Mermaid diagrams with PrivyMermaid.
Use it locally
PrivyMermaid works offline for editing and rendering. Your workspace remains on your device.
Read privacy details →Connect LLM
Use LocalEngine, Ollama, LM Studio, or another OpenAI-compatible local inference engine for private AI assistance.
Open AI settings →Frequently asked questions
Where are my diagrams stored?
Diagrams, prompts and settings are stored locally: in the extension’s local storage on Chrome and in each native app’s private local storage. PrivyMermaid does not sync them to a product cloud.
How do I use local AI in the online editor?
Sign in, open the editor, choose AI → Engine, then select LocalEngine, Ollama, LM Studio, or another OpenAI-compatible engine. Ollama commonly uses http://localhost:11434; LM Studio commonly uses http://localhost:1234/v1. For LocalEngine and other compatible servers, enter the engine’s OpenAI-compatible /v1 URL. Use “Test & find models”, choose an installed model, and save the setting. Generate, Ask and Fix requests go directly from your browser to that local engine.
If the test cannot connect, confirm the engine is running and permits cross-origin browser requests from https://diagram.kenelite.com. Local-engine settings are stored only in the signed-in browser and are not uploaded to AI Diagram.
Why is the preview not rendering?
Check your Mermaid syntax and try a built-in template to confirm the editor is working. Diagram rendering itself does not need an internet connection.
How do I export a diagram?
Use the app or extension’s export menu to save Mermaid source or an available image format. Export options vary slightly by platform.
Contact and feedback
Found a problem or have an idea? Submit feedback or report a problem. You can also email support@kenelite.com.
PrivyMermaid is a non-commercial product. We read feedback, but cannot guarantee a response time.