A collection of pro tips for making the most out of Pieces—unlock smarter workflows, richer context, and more powerful AI assistance with Long-Term Memory, Workstream Activity, and efficient navigation.
Pieces Copilot uses Long-Term Memory (LTM) to remember your work throughout the day. As you work across different applications, Pieces passively captures context from what you're doing—creating a searchable memory you can query anytime using natural language.
Think of it like having a perfect memory of everything you've worked on, researched, or discussed—ready to help you find exactly what you need, when you need it.
Released: February 2026 | Desktop: 5.0.3 | PiecesOS: 12.3.8
Powerful new capabilities for enhanced control and deeper integration:
- 🎤 LTM Audio (Preview) — Capture microphone & system audio to enrich your memories
- 🎯 Enhanced Custom Summaries — Scope by time, websites, projects & apps for precise summaries
- 📋 Custom Summary Templates — Create and save your own templates for one-click generation
- 🔌 Expanded MCP Server — Richer Long-Term Memory access in Cursor, Claude Code & more
- ⏱️ Time Breakdown Time Ranges — Pick your own time range for flexible reporting
- 📐 LaTeX Rendering — Mathematical expressions render correctly in summaries & chats
Released: January 27, 2026 | Desktop: 5.0.1 | PiecesOS: 12.3.6
Building on our 5.0.0 foundation with powerful new features:
- ⏱️ Time Breakdown — Reconstruct your billable hours from captured context
- 🔍 Timeline Search & Filters — Find anything in your history instantly
- 🎛️ Enhanced Model Selection — Pick the right model fast with inline descriptions
- 🔐 AWS Bedrock Inference Profiles — Enterprise-grade model management
- 💬 Summary to Chat — Go deeper with one tap from any summary
- 🎬 Single-Click Summary Previews — Watch video tutorials before you generate
Watch Single-Click Summary Video Tutorials | Read the full release notes →
Released: December 23, 2025 | Desktop: 5.0.0 | PiecesOS: 12.3.4
Our most significant update yet! Discover the new features:
- 🏡 A New Home Base — Unified navigation and consistent UI experience
- 🔍 Browse, Converse, Generate — Powerful in-place filtering without leaving your flow
- 👤 Personalization & Disambiguation — Smarter understanding of your work and team dynamics
- ☀️ Single-Click Summaries — Morning Brief, Day Recap, Standup, and more
- ⚙️ Behind the Scenes — Core engine improvements for better performance and accuracy
This repository contains practical guides and examples to help you get the most out of LTM queries, Workstream Activity, and navigation in Pieces.
→ Browse all guides — Full table of contents with use-case-focused overviews for every guide. General guides (LTM queries, Workstream Activity, navigation, Time Breakdown, etc.) live in guides/. The guides/MCP/ subfolder contains specialized guides for the Model Context Protocol (MCP)—connecting AI clients like Cursor, Claude Code, and Goose to your Pieces Long-Term Memory, including remote access and tool reference.
Tested, practical examples organized by use case:
- 5 Daily Work Queries: Standup generators, command line power user patterns, context restoration
- 3 Reflective Queries: Week in review, work pattern analysis, weekly highlights
- 2 Analytical Queries: Time efficiency audits, code pattern identification
- Query construction patterns and best practices
- Usage recommendations for integrating LTM into your workflow
Strategic queries for senior professionals with accumulated work history:
- 2 Practical Search Queries: Finding past conversations and research on topics
- 2 Collaboration & Efficiency Queries: Team dynamics analysis and workflow optimization
- 1 Strategic Planning Query: Quarterly performance reflection
- Query best practices and usage recommendations
- Transition guidance from 24-48 hour to 2+ month queries
A comprehensive guide covering:
- The 5 Keys to Great LTM Queries: Time, Source, Gestures, Topic, and People
- Detailed strategies for each query element with real-world examples
- Combining strategies for maximum effectiveness
- Troubleshooting tips when queries don't return what you need
- Quick reference cheat sheets for common patterns
A guide to your automatic work journal:
- How Workstream Activity works: Event capture and 20-minute summary generation
- Timeline view: Browse, search, and interact with your work history
- Interactive summaries: View, search, start chats, and share
- Pro tips for maximizing value from your workstream summaries
- Common workflows for context restoration and daily planning
Navigate efficiently in Pieces:
- What the Power Menu is and how to access it
- Quick navigation between Workstream Activity and Pieces Copilot
- Keyboard shortcuts and search tips
- Common workflows for seamless context switching
- Pro tips for staying in the flow
A video-guided walkthrough for reconstructing billable hours:
- What Time Breakdown is and why configurable time ranges matter
- Step-by-step instructions for generating a breakdown with a specific time range
- Use cases for daily timesheets, weekly billing, sprint reviews, and more
- Pro tips for matching time ranges to billing cycles and refining output
- Troubleshooting for incomplete breakdowns and missing entries
A video-guided walkthrough for capturing audio context:
- macOS permissions callout with link to the dedicated permissions guide
- Two methods to enable — Desktop App (User Profile) and PiecesOS Toolbar with platform-specific instructions
- Role-specific query examples for developers, managers, PMs, lawyers, accountants, consultants, and executives
- Privacy and control — toggle on/off anytime, local processing, no raw audio storage
A step-by-step guide to enabling LTM Audio on macOS:
- Prerequisites: Version requirements and what to update first
- Enabling LTM Audio: Two ways to activate the feature in Pieces
- Granting Microphone Access: Allow PiecesOS to capture your voice and ambient audio
- Granting Screen & System Audio Recording: Allow PiecesOS to capture system audio output
- Verification: Confirm audio context is flowing into your Long-Term Memory
- Troubleshooting: Common permission issues and how to resolve them
A video-guided walkthrough for building your own single-click summaries:
- What Custom Summary Templates are and why they save time
- Step-by-step instructions for creating, configuring, and saving a template
- 11 use case examples with copyable prompts: project updates, research roundups, client reports, meeting digests, and more
- Pro tips for naming, scoping, and maintaining your template library
- Troubleshooting for output that's too broad, too sparse, or missing templates
A video-guided walkthrough for applying discounts during checkout:
- Step-by-step instructions for the full checkout flow from the Upgrade button to purchase completion
- Finding the discount field — the green "Add discount" text above credit card fields
- Who Benefits from Upgrading? — Use cases for developers, managers, consultants, lawyers, accountants, executives, and educators
- Troubleshooting for invalid codes, missing discounts, and existing subscriptions
A simple 4-step process to update Pieces on Linux:
- Step 1 - Shutdown: Gracefully quit both Desktop App and PiecesOS
- Step 2 - Check Versions: Compare installed versions (optional)
- Step 3 - Update: Refresh Snap packages
- Step 4 - Launch: Restart PiecesOS, then Desktop App
- Troubleshooting tips for common issues
Guides for connecting AI agents to your Pieces Long-Term Memory via the Model Context Protocol. The Pieces MCP server — expanded significantly in Pieces 5.0.3 — exposes 39 tools covering full-text search, vector search, batch retrieval, temporal filtering, and cross-agent memory creation.
→ MCP Guides Index — Start here for an overview of all MCP documentation.
| Guide | What it covers |
|---|---|
| Pieces MCP and LTM Tools Reference | Complete reference for all 39 MCP tools: ask_pieces_ltm, create_pieces_memory, 14 full-text search tools, 5 vector search tools, material_identifiers, extract_temporal_range, and 16 batch snapshot tools — with parameters, examples, and agent instructions. |
| Agent Setups & Integrations | Step-by-step setup guides for 19 MCP-compatible tools: Cursor, Claude Desktop, Claude Code, VS Code, Windsurf, Goose, Cline, Continue.dev, JetBrains IDEs, Zed, GitHub Copilot, OpenAI Codex CLI, Google Gemini CLI, Amazon Q Developer, ChatGPT Developer Mode, Raycast, Rovo Dev CLI, and more. Includes transport support matrix (stdio / SSE / Streamable HTTP) and stdio-to-HTTP bridge instructions. |
| Connecting to PiecesOS via Ngrok | Expose your local PiecesOS over HTTPS using ngrok so cloud-based agents (Claude web, ChatGPT, GitHub Actions, Zapier) can reach your Long-Term Memory. Includes auto-discovery scripts for Bash and PowerShell. |
| Bridging Local MCP Clients with mcp-remote | Connect stdio-only clients (Claude Desktop JSON config, Zed, Raycast) to any remote MCP server. Covers transport strategy, OAuth 2.1, bearer token injection, and a catalog of public remote MCP servers. |
- Ready to practice? Start with the 10 example queries for 24-48 hours and try them yourself
- Been using Pieces for 2+ months? Explore the 5 strategic queries for deeper insights and performance management
- New to LTM? Dive into the comprehensive guide to understand the fundamentals
- Want to get the most out of Workstream Activity? Learn how your automatic work journal captures and summarizes your work
- Need to navigate faster? Learn the Power Menu to jump between views seamlessly
- Want to become a power user? Experiment with combining different query elements and refine your approach
- Want to connect Cursor, Claude, or Goose to your memory? Start with the MCP Guides Index and pick your tool
The best LTM queries combine these elements:
- ⏰ Time — When did it happen? ("yesterday", "last week", "in August")
- 📱 Source — Where did it happen? ("in VS Code", "from Teams", "in Chrome")
- ✋ Gestures — What were you doing? ("copied", "searched", "created")
- 🎯 Topic — What project or theme? ("customer portal authentication", "cloud security")
- 👥 People — Who were you working with? ("Sarah", "the security team")
Remember: Write naturally, like you're asking a colleague. You don't need all five elements—even one or two will get you great results.
LTM automatically captures context from:
- Code editors and IDEs (VS Code, IntelliJ, etc.)
- Web browsers (Chrome, Firefox, Safari)
- Communication tools (Teams, Slack, Discord)
- Documentation platforms (Notion, Confluence, GitHub)
- Terminal and command line
- Microphone & system audio — meetings, pair programming sessions, video calls, and presentations (enable via LTM Audio, introduced in Pieces 5.0.3)
All of this becomes searchable through natural language queries—no need to remember exact details or file names.
The best way to get good at LTM queries? Just start asking. Try different phrasings, experiment with time references, and see what works. You'll get the hang of it quickly.
For questions or support, reach out to your Pieces administrator or check out the Pieces documentation.
Unlock the full potential of your work with Pieces—learn LTM queries, Workstream Activity, and efficient navigation.