MemoryLane is a tray application for macOS and Windows whose workflow is: capture screen → summarize → local storage → detect patterns → query from any AI chat via MCP.
In terms of purpose, the tool automatically captures the screen during work, uses a user-specified model endpoint to summarize each capture into text, and then deletes the original screenshots. In the end, only the summaries, OCR text, and embedding vectors are retained, stored in a local SQLite database, with full-text search and vector search capabilities. On this basis, it can identify recurring workflows, suggest automation opportunities, and make all records queryable from any AI chat interface via MCP.
Main capabilities include: automatic screen capture, text summarization based on model endpoints, OCR text extraction, local SQLite storage, full-text and vector search, workflow pattern detection, automation opportunity discovery, and cross-AI chat querying via MCP.
In terms of usage, users can use the hosted OpenRouter endpoint by default to generate summaries; they can also point it to their own LLM, for example by running a local model through Ollama, vLLM, or LM Studio, or by using their own OpenRouter account. All configuration can be completed in settings, with no need to rebuild the application. If a local endpoint is used, captured content will not leave the user's machine. In addition, an official npm package, @deusxmachina-dev/memorylane-cli, is also provided as a CLI tool.
Applicable scenarios are mainly individuals or teams that need to review their work process, find historical screen information, discover repetitive operations, and seek automation opportunities. Through MCP integration, users can directly retrieve content recorded by MemoryLane in the AI chat they commonly use.
Limitations and notes: the official text currently does not provide specific system requirements, screenshot frequency, storage capacity, model compatibility details, performance, or pricing information; these contents are not yet provided. For privacy and permissions details, please refer to the official "Privacy & Permissions" section.
