Pieces

Pieces runs as an OS‑level service on macOS that continuously records the content you interact with—code snippets, screenshots, chat messages, and other workflow artifacts—without interrupting your tasks. It organizes these recordings into “memories” that are searchable by natural language and can be filtered by time, allowing you to retrieve a specific snippet or conversation from up to nine months of history.

The system integrates with popular development and collaboration tools such as browsers, IDEs, and chat applications through plugins, so memories are formed directly where you work. It also supplies real‑time context to large language models, including GitHub Copilot, Claude, Cursor, and Goose, enabling those models to reference your personal data while generating responses.

All data is processed locally by default, giving users full control over storage and privacy. Pieces supports multiple LLM providers and can be configured with custom keys, while its long‑term memory engine (LTM‑2) maintains the collected information for extended periods, acting as an invisible second brain for digital workers.