Pieces

On-device long-term memory layer that feeds your last nine months of work context into any LLM or IDE assistant.

Freemium· Individual free; Teams contact salesCodingMulti-model (BYO OpenAI, Anthropic, Gemini, Ollama)

Best for

Pick Pieces if you juggle Copilot, Cursor, and Claude across projects and want a single local memory that follows you between them.

Skip if

Skip it if you only need a cloud snippet manager or you're already happy letting one IDE assistant own all of your context.

Pieces is a desktop-first AI memory tool that quietly watches what you do across browsers, IDEs, terminals, and meeting apps, then exposes that context to whatever LLM or coding assistant you point at it. Its headline trick is the proprietary LTM-2 engine, which retains roughly nine months of activity locally and lets you query it temporally ("what was I doing last Tuesday on the auth refactor?") or hand it to Copilot, Claude, Cursor, or Goose via MCP so those assistants stop forgetting your project.

What sets it apart from generic snippet managers or notetakers is the local-first architecture: capture, embeddings, and storage all run on your machine by default, with optional bring-your-own-key access to OpenAI, Anthropic, Gemini, or Ollama. The free Individual tier is genuinely usable (full 9-month memory, basic copilot, all plugins), with a Teams plan gated behind sales for shared context and priority support. It's aimed at senior developers who already live inside multiple AI tools and are tired of re-pasting context between them.

Integrations are the real moat: Chrome and Edge extensions, VS Code and JetBrains plugins, an Obsidian connector, and MCP servers for the major agentic coding tools. Caveats are the usual on-device-AI ones, namely a heavy background process, a learning curve around what gets captured, and limited mobile story.

Editor's take

Pieces is the rare AI productivity tool that earns its background process by solving a real pain: AI assistants forget everything between sessions. The free tier is unusually generous and the MCP story is ahead of most competitors. We'd like clearer Teams pricing and an open-source core, but for power users it's already a daily-driver.

— The AI Tool Bible editorial team

Pros

✅ Local-first capture keeps sensitive code and meetings off third-party servers by default
✅ Nine-month LTM-2 memory bridges multiple AI assistants via MCP
✅ Bring-your-own-key support for OpenAI, Anthropic, Gemini, and local Ollama models
✅ Free Individual tier includes the full memory window and all major plugins
✅ Deep IDE and browser integration covers VS Code, JetBrains, Chrome, and Edge