Create manually, load an existing automaton, or ask local AI/OpenAI for an editable draft.
From an idea
to a system you can inspect.
Build a visual tree, teach it what matters on screen, test each decision, and run it against the exact RuneLite window you choose.
Your first run,
without guessing.
Each stage includes a check before the automaton is allowed to act.
Capture visual samples or define entity names, sensor regions, inventory, chat, and menu readings.
Connect IF / ELSE, IF FAIL, loops, variables, priorities, cooldowns, and reusable automata.
Use detection tests, dry run, live debugging, recordings, replay, and resolution calibration.
Select a RuneLite window, choose input mode, start, and use Esc as the global start/stop key.
Feature explorer
Choose a group to see what each tool is for and when to use it.
Perception
Observe the game without relying on a fixed coordinate.
Matches up to 20 visual examples anywhere in the chosen search region. Use it for icons, interface states, and messages.
Finds interactable items such as food, weapons, potions, or interface controls using visual examples or an ONNX detector.
Reads the exact visible NPC Indicators label, then clicks below the name using a configurable Y offset and optional visual fallback.
Reads the orb value or percent from a region, stores it in a variable, and lets another branch react to a threshold.
Counts slots or reads visible text so the automaton can react to supplies, messages, and contextual actions.
A real example: combat with two guardians
The main lane continues attacking. Health and prayer monitors evaluate their own conditions and only take control when their threshold is crossed.
Questions, answered
Does the AI control the game continuously?+
No. The optional AI builder converts your prompt into an editable draft. The local runtime then uses the configured recognition, logic, and actions.
Can I build everything without AI?+
Yes. The manual editor contains the same steps, branches, monitors, and testing tools. AI is only another starting point.
What is the difference between image, object, and entity?+
Image matches appearance; object represents an interactable item or interface element; entity identifies an NPC or enemy by its visible name.
Why should I hover before clicking?+
Some game interactions appear only after the cursor reaches the target. Hover plus revalidation confirms the target is still valid before the click.
Can two automata run on different windows?+
Yes, select the target RuneLite window for each run. Experimental background input reduces cursor competition, but compatibility depends on the renderer and window state.
What happens if recognition fails?+
The step can retry, run its nested IF FAIL branch, call recovery, or stop with a visible diagnostic depending on your configuration.
How do I stop everything quickly?+
Press Esc. It is polled globally, so it can stop active automata even while RuneLite has focus; from the runner it can also start again.
Are downloaded scripts ready for my screen?+
They are editable starters. Re-select regions, teach required images, confirm entity labels and interactions, then dry-run before enabling input.
Does a minimized RuneLite window keep working?+
Not always. Some renderers stop producing valid frames when minimized. Keep the client visible or use focused compatibility mode if background capture is unreliable.
Can Automata Studio guarantee account safety?+
No. Automation always carries platform-rule and account risk. Review every flow and decide where it is appropriate to use.
Download Free or choose an editable script to learn from a real example.