$ loading_
Store and retrieve cross-session UI visual memory for computer-use agents.
Copy the install command and let the AI configure it · recommended for beginners
No copy-paste install info for "Polaroid" yet — see the docs or source repo.
Save the current webpage screenshot as visual memory nodes labeled 'settings button, user avatar, left navigation'. Next time I say 'click settings', first retrieve the matching region from historical visual memory, then perform the action.
The tool stores semantic scene nodes and later locates the target control region by intent.
During the expense submission workflow, continuously store each page screenshot as semantic scenes. If layout changes break normal selectors, retrieve historical visual memory by intents like 'submit expense', 'upload attachment', and 'confirmation dialog' to continue the flow.
Creates reusable visual memory that helps automation continue even after UI layout changes.
Query the UI memories related to 'export report' from the last three sessions, and return the pages involved, button locations, and nearby elements so I can analyze why the agent failed.
Returns an intent-based summary of historical UI scenes to help diagnose agent failures.
Let AI control desktop apps for automation, testing, and repetitive tasks.
Let AI agents visually test frontend apps with screenshots, clicks, and logs.
Give AI agents offline adaptive memory that retains what matters most.
Lets AI understand and operate web interfaces through unified UI scene graphs.
Provide episodic memory and architectural policy management for AI agents.
Give AI agents long-term memory with user-scoped storage, recall, and deletion.