Describe your macro.
Build it with AI.
Connect your assistant to JAQ Macro through MCP. It can inspect your project and turn your request into editable steps, conditions and reusable groups. Review the result in Windows before you run it.
Describe it in chat, edit it in JAQ
“The app is open. Find Ready, click it once, and verify that the next window appears. If you cannot find it, stop after three attempts.”
- Connect
Connect a compatible client such as Codex, Claude Code, Kiro or Antigravity through Account → MCP.
- Build the workflow
The assistant inspects your project, reuses visual references, and adds editable steps, conditions and bounded retries.
- Test it locally
Review the steps in the editor, test a group and press Start in JAQ Macro when ready.
From pixels to targets.
From targets to workflows.
One connected workspace.
Example selection, background removal, training and live tracking.
Explore the real tracking studio ↗Examples to instant tracking
Start VOT from an annotated target and Alpha mask. A separate YOLO training run is not required for the first trial.
YOLO11 + VOT hybrid
Combine appropriate-model detection on DirectML/CPU with local visual tracking.
360° data preparation
Build your dataset with GrabCut, Copy-Paste, rotation, shadows and appearance settings.
This original world was created specifically for JAQ. Matches come from the real local detector; this is not a commercial-game performance or compatibility benchmark.
Think beyond the click.
Design what happens next.
A button does not always stay in the same place. A window does not always load at the same speed. Observe the actual state, evaluate the result and define the next action.
Capture what matters.
Find targets with images, trained objects, colors and OCR.
Explore visual search ↗Give each result a path.
Respond to different states with conditions, loops and formulas.
Explore conditions ↗Work in the right place.
Connect mouse, keyboard, window and program actions in meaningful groups.
Learn workflow architecture ↗Turn what you see
into what happens.
Go beyond fixed coordinates. Locate a target, read its text and store the result. Let the actual state guide your next action.
Open the object-library guide ↗Small building blocks.
Powerful workflows.
Window
Find the right application, check its state and attach the workflow to its window. The starting point for automation across multiple programs.
Find image
Search for a fixed image, trained object, pixel color or color coverage. Assign different behavior to found and not-found outcomes.
Read text / OCR
Read a selected area, filter the output and store it in variables. Use local engines or choose Gemini/OpenRouter for suitable OCR tasks.
Navigate to coordinates
Follow world coordinates with local OCR, or replay a recorded route. Choose movement and heading methods for the target application.
Condition / If
Branch on a previous result, a built-in window or file check, or a formula. Design both the true and false paths.
Keyboard and text
Combine keys, combinations, literal text and OCR results in an ordered input list. Tune each item’s behavior and delay.
Choose an example.
Adapt it to your task.
Read your position.
Adjust your heading.
Design feedback-driven movement with local OCR, direction arrows, the actual key behavior, tolerance and waypoints. Use route replay and checkpoints as a separate method.
Explore coordinate navigation ↗Control is in the details.
Before you begin.
Is JAQ Macro only for games?
No. Window, program, file, mouse, keyboard, image and OCR modules can work together for compatible Windows applications. Test the target’s input and capture support.
Do I need to write code?
Build flows with visual modules. Add numeric, text or boolean expressions when useful, or ask an MCP-capable assistant to help design the logic.
Is AI required?
No. The local editor and local image/OCR methods can be used independently. MCP-assisted design and Gemini/OpenRouter AI OCR are separate options.
Will background mode work in every application?
Background mode depends on window-message support and accepts limited input types. It is not guaranteed for every application, including Raw Input/DirectInput games.
Can I move a flow with its images?
The relevant Pro right enables .jaqb export of selected groups and linked PNG templates. Check local trained-model requirements separately, then verify targets and geometry on the other computer.
See the difference
in the workflow you build.
What happens when the target changes? Which path handles missing data? Can multiple groups share one check? Compare approaches through actual needs.
Compare capabilities side by side ↗All features →