What is PiecesOS? | Pieces Docs

Learn about PiecesOS, the background service that powers Long-Term Memory, on-device capture, and MCP integrations across the Pieces ecosystem.

**Action required:** Cloud services no longer work on PiecesOS versions prior to 12.4.0 (deprecated June 5th 2026). Update to 12.4.0 or later to restore Chat, Work Summaries, and Single Click Summaries.

What is PiecesOS?

PiecesOS is a background service that runs on your machine. It orchestrates on-device data capture and storage, coordinates AI model requests, and serves as the bridge between your workflow and every Pieces product—including the Pieces Desktop App, MCP integrations, and the CLI.

What PiecesOS Does

PiecesOS powers three core capabilities:

Agentic Long-Term Memory (LTM-2.7)

The LTM-2.7 Engine continuously captures workflow context—code you copy, screens you view, audio you hear—and stores it locally on your device. This memory powers Timeline, Conversational Search (Agentic Chats), and Single-Click Summaries (Agentic Summaries) with real context from your day.

The agent reasons across your memory in multiple turns, following threads, cross-referencing context, and building complete answers instead of one-shot guesses. It can search your memories, the web, your calendar, local files, and browser history automatically.

AI Models

PiecesOS coordinates AI model requests for all Pieces products. Choose from four model families—Claude, Gemini, Grok, and ChatGPT—each available in Fast, Balanced, and Extra Thinking modes. See Choose a Model to pick a family and mode.

MCP Support

The Model Context Protocol (MCP) is an open framework that lets LLMs access your workflow context. PiecesOS serves as the MCP host, connecting tools like Cursor, VS Code, Claude, and ChatGPT to your Long-Term Memory without custom integrations.

Pieces MCP integration with Cursor for context-aware changes

Pieces MCP integration with Cursor showing context-aware documentation changes

Privacy & Local-First Design

All data captured by PiecesOS is stored locally on your device. Capture, indexing, and storage happen on-device, and PiecesOS applies on-device ML to filter out sensitive information and secrets.

AI features that need a large language model run in the cloud, where only the scoped context for that request is sent—the rest of your data stays on your machine. We're **SOC 2 Type II** certified and never use your data to train models. [Learn more about privacy and on-device storage.](/products/core-dependencies/on-device-storage)

Installing PiecesOS

PiecesOS installs automatically with the Pieces Desktop App. If you want to run it standalone (for example, with MCP integrations only), see Manual Installation.

Explore PiecesOS

Learn about Agentic Long-Term Memory, the agent toolbox, and how to enable, pause, and control memory capture. Manage PiecesOS settings, LTM, MCP, ML processing, and updates from the menu bar or system tray. Where Pieces stores your data locally, how to find logs, and how to back up or reset your database. Fix common installation issues, check system specs, update PiecesOS, and find logs on macOS, Windows, and Linux.