Understand Context
In AI, context is the working memory a model can reference when it generates a response: instructions, conversation history, documents, tool outputs, and other relevant signals. AirJelly extends that idea from a single chat window to your desktop workflow, so the context behind your work can be remembered and reused later.
What Is Context?
For an AI system, context is the information available at the moment it reasons or responds. It can include your current request, previous messages, relevant documents, tool results, system instructions, and other signals that help the model understand what is happening. In AirJelly, that concept becomes a personal working-memory layer across your apps: the product can connect what you saw, typed, discussed, and returned to later so you do not have to rebuild the background from scratch every time.
Why Can AirJelly Capture Your Context?
After you grant Screen Recording and Accessibility permissions, AirJelly can use a screenshot-based capture mechanism to understand your visible workspace. Each time you press Return or hit Enter, AirJelly captures the screen state, reads the current information, and stores the extracted signals as long-term context. Later, when you use AirJelly, Record Memory can retrieve and reuse that context so the product can answer, suggest, or act with awareness of what happened before.
How Context Becomes Usable
AirJelly uses visual understanding and accessibility signals to extract the app, window, current action, written content, goal, topic, project, and people involved. Those structured results can then appear in Timeline, Memory, Agent, and proactive suggestions. When you ask Agent, write a prompt, or build an automation, AirJelly can use a summary of what happened around the task instead of a pile of isolated screenshots.
How To Stop AirJelly From Capturing Certain Context
When you press the Option key twice in a row, AirJelly covers its eyes. During that animation, it does not take screenshots, record, or observe anything on your screen. Use this whenever you are in a private moment, handling sensitive content, or looking at something you do not want to become reusable context.
