Announcing Pieces Long-Term Memory agent (LTM-2): the AI that remembers everything you work on
Introducing Pieces LTM-2: The first AI with long-term memory that remembers everything you work on. Boost productivity with seamless recall!
I’m Tsavo Knott, CEO and Technical Co-Founder of Pieces, and today, we're excited to introduce Pieces Long-Term Memory Agent (LTM-2) — the first AI that remembers everything you work on, across all your tools and workflows.
The missing piece in AI: memory
AI has transformed software development, but it’s always lacked one critical component: Memory.
Every day, we create and lose valuable data—code snippets, insights from conversations, research findings—scattered across applications, buried in histories, or simply forgotten.
While LLMs act as a “prefrontal cortex” for AI, generating content, almost no one is building the digital hippocampus – the memory system that stores, retrieves, and recalls critical context over time.
That’s where Pieces comes in.
The memory problem in your workflow
Think about your workday:
How many tools do you switch between? Research shows developers transition between apps over 1,000 times a day, losing time and context with each switch.
Your digital tools forget everything—unlike your brain, they don’t preserve interconnected context and temporal relationships that shape your work.
Introducing LTM-2: The First OS-Level Long-Term Memory Agent
Pieces captures, structures, and resurfaces your work across macOS, Windows, and Linux, making memory an integral part of your operating system.
☑️ 9 Months of On-Device Memory – Your code, notes, links, and discussions are preserved automatically, letting you pick up where you left off.
☑️ Interactive Workstream Activity – See what you did, when, why, and how in a full interactive timeline.
☑️ Natural Memory Retrieval – Ask questions like:
"When was the last time I ran into an error like this?"
"What projects was I working on around Thanksgiving?"
"Can you generate a stand-up note summarizing everything I worked on this week?"
Get instant, contextual answers with full recall.
Our continuous hierarchical summarization technology structures your work like a human brain, consolidating short-term interactions into long-term knowledge at multiple levels of detail.
Private, performant, and built for teams
☑️ 90% of the system runs offline, with only a small fraction requiring cloud-based LLMs (soon transitioning to hyper-efficient on-device transformers and SLMs).
☑️ Full user control – Pause or turn off memory formation, block specific apps, and delete data by source, topic, or time range.
☑️ Effortless team sharing – Share long-term memory context across teams, ensuring collaboration without losing knowledge.
Together, we’re building OS-Level artificial memory
LTM-2 isn’t just a product launch—it’s the beginning of a new category: OS-Level Artificial Memory for developers and digital professionals.
LTM-2.5 and LTM-3 are already in the works, pushing the boundaries of AI memory and context-aware workflows.
This journey wouldn’t have been possible without our community, team, investors, collaborators, and supporters—thank you for believing in this vision.
We can’t wait for you to experience how Pieces changes the way you work by remembering so you don’t have to.
Try Pieces LTM-2 now and take control of your digital memory!
If LTM-2 sounds interesting to you, we would love your support on our Product Hunt Launch!
Cheers,
Tsavo Knott & The Pieces Team
