Botream documentation
Botream is an on-premise RPA platform: you build automations in the browser, and robots on your Windows machines run them against web applications, desktop programs, files, mail, databases and APIs.
How Botream fits together
| Part | What it does | Where it runs |
|---|---|---|
| Botream Server | The web app and its API, the scheduler, work queues, vaults, the audit log. Keeps everything in a SQL Server database. | A Windows server, as the BotreamServer service |
| Web app | The task editor, runs and their logs, schedules, queues, vaults, administration. | Any modern browser (served by the server) |
| Botream Agent | The robot: receives tasks from the server and runs them; also captures elements and records what you do. | Each robot machine, as a tray app (with a desktop) or the BotreamAgent Windows service |
| Botream Element Capture | The Chrome/Edge extension that captures web elements and records clicks and typing in your browser. | The browsers you build tasks in |
| SQL Server | The only external component: no Redis, no message broker. | Your database server or the server machine |
A task is a list of commands. You put it together in the editor — as blocks, or as text in the task language — and the server sends it to an agent, which runs it and streams its log back live.
browser --action:open --url:"https://shop.example.com" --output:shop
capture --browser:$shop$ --action:settext --dompath:"#search" --value:$customer$
capture --browser:$shop$ --action:click --dompath:"//button[@type='submit']"
capture --browser:$shop$ --action:gettext --dompath:".order-total" --output:total
log --message:"Total for $customer$: $total$"Start here
- New installation: Install the server, then Install agents and the browser extension.
- First automation: Quick start walks you from an empty task to a finished run.
- Building tasks: The task editor, the task language and variables, then web and desktop automation.
- Faster ways to build: Recording your own clicks, and the AI features — the AI Task Generator can even work in a browser itself and capture the elements as it goes.
- Every command: the command reference, generated from the same definitions the editor and the robots use.
What you can automate
- Web applications in Chrome, Edge or Firefox through Playwright: clicks, forms, tables, downloads, tabs, frames — no ChromeDriver to keep up to date.
- Windows programs through UI Automation: buttons, fields, menus, grids; keys, mouse, screenshots, OCR and picture matching for anything else.
- SAP through SAP GUI Scripting: sign in, transactions, fields, ALV grids and the status bar.
- Data and files: Excel, CSV, JSON, XML, PDF, databases, files, folders and ZIP archives.
- Communication: e-mail (SMTP and IMAP, OAuth2 included) and REST APIs.
- Unstructured input with the
aicommand: answers, fields pulled out of documents and e-mails, classification.
Built for reliable operation
- Selectors that keep working: fallback locators, anchors, self-healing with a similarity threshold, and selector health that reports elements drifting before they break.
- Clear failures: every failed run keeps the failing line, the variables and a screenshot of the robot's screen, and the AI can explain what happened.
- Operations: schedules, event triggers (files, mail, webhooks, hotkeys), global values, work queues, agent groups, performance measurement and a complete audit log.
- Security: passwords only in encrypted vaults, role-based access, every security-sensitive action audited, and everything on your own network — including the AI model if you run one in-house.