AI memory for agents · Project handoffs

Pick up the thread, not the pieces.

A fresh session should build on the work you already did. We bring the original request, recorded progress and results together, so you can see what matters and give your next agent a useful starting point.

Claude-Mem mascot carrying a useful project note

AI memory in production

70 days of work. One searchable memory.

Alessandro used our open-source engine across three machines for 70 days. His customized deployment captured the decisions, fixes and discoveries behind the work—and published the data.

Observations captured
22,718
Bug fixes recorded
1,262
Decisions recorded
757

See how he made that knowledge searchable. Then give your own agents a memory to build on.

Claude-Mem’s mascot files a note in a shared card catalogue connected to three computers.
Alessandro’s published field study · March–May 2026

Reviewed November 18, 2026 · By Claude-Mem

01

Recover the goal behind the work.

Open My Work to find the original request, recorded follow-ups and latest result in one place. Filter by project or search for the work you want to resume, then inspect the source records behind it.

02

Keep the result and next move connected.

See what came out of the request and what still needs attention. Eligible local deliverables appear alongside the work that produced them, helping you review the artifact with its context instead of searching through disconnected chats.

03

Give the next agent a clear handoff.

Copy handoff gathers the goal, recorded state, next step and recent updates. Bring that text into the agent session you choose, with the current repository and required files, and continue from the actual state of the work.

04

Build continuity into your review.

Check the latest evidence before continuing. My Work records your review of a result version and lets you undo it. When a substantive source update changes the result, the earlier check is invalidated so you can review the new version.

Questions before you start

How do I open My Work?

Install Claude-Mem and open the worker URL printed by the installer. My Work is the default local viewer and reads saved local data without a cloud account, a new model call or an internet connection. Use Memory & settings to explore the wider saved history.

Does Copy handoff send work to another agent?

It copies text for you to use in the session you choose; it does not send a message, assign work or launch an agent. If clipboard access is unavailable, select the displayed handoff text.

Will the next device have all my files?

Cloud Sync replicates eligible memory records, not your repository or arbitrary local deliverable files. The deliverables bucket stays on the computer that collected it. Make the current checkout and required artifacts available to the receiving session.

What do the work statuses tell me?

In progress reflects the latest saved state, not live agent presence. Ready to check means the parent request was reported complete; a finished subtask does not finish the parent. Mark checked records your review of that version, not independent proof of correctness or deployment. A substantive source update invalidates an older check.

Can I share this work with a team?

TeamBrain is the separate shared-work view with its own membership and project-sharing setup. This page covers resuming your own requests. Cross-device shared My Work state requires compatible readers, a supporting hub and its explicit shared-state setting; local review notes and files have their own scope.

Go deeper

Explore the source documentation and continue with the topic that matches your work.

Give your next session the context it needs.

Install Claude-Mem and keep your goal, progress and next step connected. Add CMEM Pro when you want eligible memory across your devices.