Quickstart

This walkthrough takes you from an empty workspace to a published, API-callable workflow. All you need is a task you would normally do in a browser: looking something up, filling a form, or downloading a report. We will use a small public site, books.toscrape.com, so you can follow along exactly without any accounts or credentials.

There is nothing to install and nothing to configure. RamAIn launches a cloud browser for you, so your machine and your own browser stay untouched.

Step 1

Create a workflow

Sign in and you land on Workflows, your team's library of automations. Every workflow shows its status (Published or Draft), its last run, its triggers, and how many blocks it contains. The banner at the top shows your team inbox address; you will meet that address again when you set up email triggers later.

Click New Workflow. The workbench opens with a chat panel and launches a dedicated cloud browser for this session. The browser is real (not a simulation or a recording tool), it runs in RamAIn's cloud, and it is ready in a few seconds. You will see it live on the left side of the screen for the whole session.

Step 2

Teach the agent

Type what you want done into the chat, the way you would brief a colleague:

Go to books.toscrape.com, open the Travel category, and read out the title and price of the first three books.

Then watch. The agent navigates to the site, finds the Travel category in the sidebar, opens it, and reads the first three entries. Every action is visible in the live browser, and every action is narrated in the chat, so you always know what it is doing and why. When it finishes, it reports the titles and prices back to you in the chat.

This conversation is the whole "programming" step. You never write a script or record clicks. If the agent takes a wrong turn, interrupt it and correct it in plain English. If it is unsure about something, it asks you before proceeding. If a page needs a human (a login it does not have, or a choice only you can make), it hands you control of the browser and waits.

Tips for a good first session

• Give one task at a time. You can chain more instructions as it finishes each one.
• If a login is needed, the agent asks for credentials with a secure form. They are saved to Passwords and never echoed in chat.
• Mention what should be reusable. Saying "the order number will be different every time" tells the compiler to make it an input.

Step 3

Save, and the session becomes a workflow

Rename the workflow using the title field in the top-left (something like "Travel book prices"), then click Save. This is the moment RamAIn turns a one-off session into a reusable automation. The compiler reads the whole session, every step and every page the agent visited, and produces a workflow you can read at a glance on the Editor canvas: an Input block, the agent steps that do the work, and a Parser that structures the results.

Notice what the compiler did on its own: it spotted that "three books" was a choice rather than a constant, and turned it into a num_books input you can change on every launch. Ask for five books next time without touching the workflow itself.

This is also where you refine things. The Editor is a chat, just like the Builder. Say "also capture the star rating" or "if the category has fewer books than requested, return what exists and note it", and the Editor updates the graph in front of you. For everything you can do here, see the Edit phase.

Step 4

Publish

Click Publish. Until now your workflow was a Draft: visible to your team, editable, testable, but not yet wired into the outside world. Publishing promotes it to Published, which means three things at once:

• it appears with a Launch button on the Workflows page for everyone on your team,
• its email and schedule triggers go live,
• and it is exposed as a REST endpoint you can call from your own systems.

You can keep editing after publishing. Edits accumulate as a new draft version, and you decide when the changes go live, so a half-finished edit never affects the version your triggers and API callers are using.

Step 5

Launch it

Back on the Workflows page, hit Launch. The workbench opens on a live view of the run, and the agent asks for any inputs it needs right in the chat; ours has just one, num_books. Answer it and the run proceeds. You can watch it live, step by step, exactly as you watched the original session, or close the view and let it finish on its own. Either way the run and its outputs land on the Sessions page.

Manual launches are only one of four ways to start a workflow. Head to Triggers to put it on a schedule ("every weekday at 7am"), wire it to your team inbox so an incoming email starts a run, or grab the ready-made API snippet and launch it with a single POST from your own code.

Where to go next

You have now seen the full lifecycle: build by demonstration, save to compile, publish, and launch. The rest of the docs go deeper on each phase.

The Build phase

Everything you can tell the agent and everything it can do while teaching a workflow.

The Edit phase

The blocks, the node inspector, and every tool in the editor window.

Launching & triggers

Schedules, email triggers, and the API, with examples for each.

Agents

Compose this workflow with APIs, code, and loops into a bigger automation.