AI chat on an infinite canvas
NodePad lets you think in every direction. Talk to frontier models like you always do, ask side questions without derailing anything, and pull your best threads into one answer you can act on.
No credit card to start. Your first workspace is already waiting.
Linear chat loses your thinking.
Every good AI conversation branches: a tangent worth chasing, a second opinion, a version worth keeping. Chat buries all of that in one scrollback. A canvas keeps it in view, so you spend your time thinking instead of scrolling.
Three moves you can't do in a chat window.
Every message on a NodePad canvas is a node you can build on, and the model only ever sees the context you chose. No hidden prompts, no injected instructions. That is what makes these three moves possible.
Fork
Explore the tangent on its own branch. Your main thread stays clean, and both paths keep the full story.
@-Reference
Pull the exact message you need into what you're writing, even from another branch. No re-explaining, no copy-paste.
Merge
Run several chats on one canvas, then ask one question with all of them in scope. One answer, informed by everything you learned.
Built for work that doesn't fit a single line.
Brainstorm
Try three directions at once, with different framings or different models, and walk away with one plan you actually believe in.
Research
Sources, competing threads, and the synthesis stay in one workspace. When you write it up, everything you need is already on screen.
Debug
Give each hypothesis its own branch. The fix stays connected to the evidence, so you can defend it later.
Every model. Every plan.
Every plan reaches every model. Bigger plans just give you more room to work each month.
Cancel anytime. Your workspaces stay yours to browse and export.
Run NodePad on your infrastructure.
Self-hosted or fully air-gapped. OIDC SSO, SCIM, audit logs, and bring-your-own model keys or inference endpoints.
Questions people ask us.
What is NodePad?
NodePad is a workspace for thinking and working with AI on a canvas instead of in a chat. Every message is a node you can move, fork, reference, and merge, so a single workspace can hold a debugging session, three competing research threads, and the synthesis that came out of them. It's chat on a canvas, and it's also agents on a canvas: Claude Code plugins (MCP tools and Agent Skills) work out of the box, and you can bring your own plugin marketplaces.
How is this different from ChatGPT or Claude.ai?
ChatGPT is a linear chat. NodePad is a canvas. You fork threads, run multiple models in parallel, and keep every related conversation in one workspace instead of scattered tabs.
Which models does NodePad support?
All frontier models are supported: Claude, GPT, and Gemini. Enterprise plans additionally support any OpenAI-, Anthropic-, or Gemini-compatible endpoint so you can route calls through your own inference (Ollama, vLLM, Bedrock, Vertex, Azure OpenAI).
Can I self-host?
Yes. Enterprise customers run NodePad on their own infrastructure, including fully offline / air-gapped deployments. See the Enterprise page for options.
Does NodePad change what the model sees?
No. The model receives exactly the context you built: the thread you are on, plus anything you referenced or merged in. There are no hidden system prompts and no injected instructions. That control is the point.
How does pricing work?
Plans start at $5/month. Every plan reaches every model. Go, Pro, and Max differ only in how much included monthly usage you get, and you can add extra usage anytime. See the pricing page for details.
What happens to my data?
Your workspaces are private to your account, and your prompts and responses are never used to train our models. See the privacy policy for the full picture.
More questions? Read the full FAQ or email us .
Your next hard problem deserves a canvas.
Open a workspace and ask the question you keep coming back to.
you draft Ask the question you keep coming back to claude-opus-4-8Plans from $5/month. No training on your data. No hidden prompts. Cancel anytime.