Documentation
A canvas where your notes, files and drawings are objects you can draw between. Claude and ChatGPT read and write the same graph over MCP.
The model
Everything in ideagraphs is an Object. A document is an Object, a drawing is an Object, an uploaded file is an Object, and so is a person, a type and a space. The same few operations read and write all of it, which is why there are no type-specific tools and never will be: the types are yours to invent.
Edges between Objects are typed, and a space is an Object that selects and presents others. A drawing is a body of nodes rather than a separate kind of thing. Objects you make stay on your device until you sign in.
Connecting an assistant
ideagraphs runs as an MCP server, so Claude or ChatGPT reads and writes the same graph you see on the canvas, in the same space you are working in. Authentication is OAuth 2.1 with PKCE and dynamic client registration, so connecting is a consent screen rather than a pasted API key.