Getting started
Flow is where you assemble a bot, or an automated piece of work, and hand it to people: as a chat page behind a link, a widget on your own site, a Telegram bot, a call from another system, or a run on a schedule. You assemble it with the mouse, out of cards. No programming.
This page is the map of the start: what you need, what happens in the first five minutes and where to go next. The step-by-step build with a result at the end is the first tutorial.
What you need
An account, and nothing else. Signing up creates a workspace and a free plan for you — no model key, no card, no installation. Models come from the platform catalog, so your first bot answers as soon as you pick one from the dropdown.
The first five minutes
-
Sign up. You land in Workflows, and it is empty — that is expected: No workflows yet.
-
Press New Workflow. The dialog has a Workflow name field and, below it, or start from a template.
-
Take a template. It is the shortest path to a working graph: the whole workflow opens in the editor and all you do is pick a model. An empty workflow opens as an empty canvas — there is no entry and no exit on it, and you drag both in from the palette yourself.
-
Pick a model on the LLM node and press Save.
-
Switch to the Test tab and write the bot a message. There is no separate run button in the editor — a message in the chat is what starts the workflow.
-
The Deploy tab → the Public chat channel → the Deploy version 1 button. You get a link you can open in another browser or send to a colleague.
Workflows
Workflows
No workflows yet
| Name | Status | Runs 7d |
|---|---|---|
| — | — | — |
What it costs
A credit is a cent, and it is spent on work actually done. One chat turn adds up like this:
1 (the run itself)
+ 1 (the LLM node call)
+ input/1000 × k_in + output/1000 × k_outSo two credits minimum per answer, however short the reply. Everything above that is the model’s rate: a turn of 2,000 input and 500 output tokens costs about 2.1 credits on an inexpensive model, around 4 on a common strong one, and tens of credits on the expensive reasoning models (the priciest one in the catalogue passes a hundred credits for a single turn). The free plan grants 100 credits a month — roughly forty answers on an inexpensive model, about twenty on a strong one, and less than one on the most expensive.
Every model’s rate is shown in Settings → Models, next to its name. The full breakdown of what is charged is in Credits and cost.
The edges of the free plan
Worth knowing up front — otherwise you meet them as a refusal at the moment everything is already built.
| What | On the free plan |
|---|---|
| Credits | 100 a month, no overage |
| Deploy channels | Public chat and the widget only; Telegram, webhook and schedule are paid |
| Deployed at once | one workflow |
| Workflows in total | five |
| Knowledge bases | two |
| Your own model (your key) | not allowed |
Exact values for every limit are in Plan limits: that table is generated from the platform’s own settings and cannot drift from reality.
Where to go next
Your first chatbot
Step by step: build it, test it and put it behind a link in one sitting.
What Flow is
The product model in five minutes: graph, nodes, wires, channels.
Examples
Ready-made graphs — copy the nodes into your own project.
Glossary
Short definitions of the terms used in the app and the docs.