Petrograph

A diagramming application for creating mind and organizational maps to help AI understand the big picture. Exports and imports JSON for AI integration.

The problem

We were paying $1,000 a year for Miro to create diagrams that we'd then have to manually describe to AI tools. The AI couldn't see the diagram. It couldn't understand the relationships between nodes. Every time we wanted AI to work with our system architecture or org charts, we had to write paragraphs explaining what the diagram showed.

What we built

A diagramming tool that exports its structure as JSON. Feed that JSON to any AI and it instantly understands the relationships, hierarchy, and connections in your diagram. Import AI-generated JSON back and the diagram updates automatically.

Built in one day. Cancelled our Miro subscription the next morning.

Screenshots

Features

What it does

Visual diagram editor

Canvas-based editor with drag-and-drop nodes, connections, grouping, and color coding.

JSON export for AI

Export diagram structure as JSON that any LLM can parse and reason about.

Bidirectional sync

Import AI-modified JSON back and see changes visualized instantly.

Multiple diagram types

Mind maps, org charts, system architecture, and flow diagrams.

Collaboration ready

Share diagrams with team members for real-time editing.

Process

How we built it

Identified the AI-diagram gap

Documented every time we had to manually describe a diagram to an AI tool. It was happening daily.

Designed the JSON schema

Created a node-edge-metadata format that any LLM can parse and reason about without special training.

Built the visual editor

Canvas-based editor with drag-and-drop nodes, connection lines, grouping, and color coding.

Added bidirectional JSON

Export diagram as JSON, feed to AI, get modified JSON back, import and see the changes visualized.

Tested with real architecture diagrams

Used it to diagram three client system architectures. AI understood and suggested improvements from the JSON alone.

Reflections

What we took away from this project.

What went wrong

The first JSON schema was too verbose. AI models would hit context limits on large diagrams. We had to compress the format significantly.

What went right

The bidirectional flow was the breakthrough. Being able to ask AI "reorganize this org chart" and see the result instantly changed how we work.

What we learned

The gap between visual tools and AI tools is massive. Any tool that bridges that gap has immediate value. People don't want to describe what they can see.

Want something like this for your business?

Tell us what you're working on. We'll tell you how fast we can build it.