Build and run your first AI workflow in about ten minutes. Knitch runs in your browser, so there is nothing to install.
A workflow is a set of nodes wired together on a canvas. Each node does one job: take input, call a model, run code, branch, send an email. You connect them, run it, and Knitch executes the graph on durable infrastructure, streaming each node's output as it goes and tracking what every step costs.
From Workflows, click New. You'll get an empty canvas. Give it a name in the top bar.
Open the node menu with Cmd K, or right-click the canvas. Add an Input node and an AI node. Drag from the Input's handle to the AI node to connect them.
Select the AI node to open its panel. Pick a provider and model (Anthropic, OpenAI, Google, xAI, and more, kept current from the model registry), then write a prompt.
To feed it the upstream value, reference another node with {{Node Label.field}}. Type @ in any text field to pick the node and field from a list, so you don't have to remember the names.
Add an Output node and connect the AI node to it. Save with Cmd S, then run. Each node lights up as it executes, with the model's output streaming in live.
Open the run to see the output along with per-node and per-run cost and timing. Every run is saved in history, so you can compare runs and see exactly where time and money went.
Swap manual runs for a trigger. A webhook fires the workflow when something posts to a URL, a schedule runs it on a cron, and a mailhook runs it when an email arrives. Now the workflow reacts to events without you in the loop.