Grafy

Graph Chat — the branching workspace

The flagship: prompts, images, video, text, audio and 3D as nodes in a graph you can branch, compare and merge — plus PDF mind maps.

Graph Chat on Grafy
Sign inExplore Grafy

How it works · From the blog · All tutorials · Community gallery

What it makes

Graph Chat turns a conversation into a document you can branch. Every prompt, image, video, clip of audio and 3D model is a node on a canvas, so a second idea starts a second branch instead of overwriting the first. Drop a PDF in and it becomes a mind map of the same shape.

For work that does not go in a straight line: research with three hypotheses, a campaign with four treatments, a document you keep interrogating from different angles.

How a session works

  1. Open a graph, type the first prompt, and pick the model and the kind of answer you want back.
  2. Branch from any node to try a different wording, a different model or a different medium beside the original.
  3. Merge the branches you kept into one answer, or export the graph and take the whole thread with you.

What it costs

Editing the canvas, branching and comparing cost nothing. Only the generations you run spend credits, and free-to-serve models keep working when the balance is gone.

Generation spends credits: paid plans include a monthly allowance, the Free plan starts with a one-time welcome grant, and you can top up at any time.

Included from the Starter plan.

Examples from the community

  • Rover
  • Double exposure portrait, a profile silhouette filled with a misty pine forest at dawn, clean white background outside t
  • Backlit portrait at golden hour, sun flaring through hair creating a rim of light, warm haze, shallow depth of field wit

Published by people using Grafy. The gallery has the rest. Community gallery

Questions

What is a graph workspace?
A canvas where each prompt and each answer is a node you can point at. Instead of one scrolling thread, you get a shape you can branch, compare side by side and merge back together.
Can I use it on a PDF?
Yes. A document you drop in is read into a graph of its own sections, so you can question one part of it without losing the rest, and keep the answers attached to where they came from.
Which models can I use?
The catalogue covers local and cloud models for text, images, video, audio and 3D. You choose per node, so one branch can answer with a fast model and the next with a careful one.
Does my work stay private?
A graph is yours and nobody else can open it. Publishing to the community gallery is a separate, deliberate step you take on one result at a time.