Short, beginner-friendly walkthroughs. Each plays a real workflow on the canvas and zooms into every node so you can see what it does and what to fill in.
Build your first workflow
A complete AI pipeline, end to end: type a prompt, an LLM rewrites it into something richer, and a Text To Image node turns it into a picture — all from connected nodes, no code.
You’ll learn how nodes pass data through their handles, how to read live status (running rings, streaming text, the progress bar), and where generated output appears on the canvas.
Connect & run
The core loop, one step at a time. Add a node, wire its output into the next node’s input, press Run, and read the result.
You’ll learn what inputs, outputs, and handles are, how to run a graph and watch nodes complete, and how to find a node’s result in a Preview.
Generate a list
One prompt, many results. A single LLM node turns a topic into a structured list, streaming each item as it arrives — the pattern behind batching and loops.
You’ll learn how to drive an LLM node from an input, watch multi-item output stream in, and pass a list into the rest of a workflow.
Ask the AI
The simplest chat-style graph. Type a question, send it to an LLM node, and watch the answer stream in phrase by phrase before it lands in a Preview.
You’ll learn how to feed a question into an LLM node, watch the answer stream as it generates, and reuse it downstream.
Combine two inputs
The first graph that branches in. Two text inputs flow into one Format Text node that fills a template, composing a single result from reusable parts.
You’ll learn how to wire several inputs into one node, compose text with ``, and build prompts from reusable parts.
Summarize a document
Long text in, key points out. A single Summarizer node condenses an article, transcript, or any block of text into a short summary, streaming it as it writes — the pattern behind the Meeting Transcript Summarizer example.
You’ll learn how to feed a long passage into a Summarizer, watch the summary stream as it generates, and pass the result into the rest of a workflow.
Describe an image
The first multimodal graph. Drop a picture into an Image Input, wire it into an Agent, and watch the model look at the image and describe it in words — the vision pattern behind captioning and alt-text workflows.
You’ll learn how to bring an image into a graph, send it to a vision model, and reuse the streamed description in any downstream text node.
Ask the chat agent
A different surface: Chat. A question goes straight to the agent, which calls a web-search tool in the open — no hidden spinner — then streams its answer back token by token.
You’ll learn how to send a message from Chat, watch a tool call run in the open, and read a streamed answer as it arrives.
Cut a scene together
The timeline (video) editor: trim a clip, drag in another shot, drop in a caption that syncs word-by-word with the audio, then scrub the finished cut.
You’ll learn how to trim and arrange clips on real tracks, add a caption synced to the audio, and scrub and preview a cut without leaving the browser.
Ready to build your own? Start with Quick Start, then browse the Examples.