The day's work, not redone at night
Many agents, one playbookSolo builder with two machines
Work started on the office laptop and continued on the home desktop. The desktop's agent knew nothing about the day.
How do I stop my home desktop's AI agent redoing what my laptop's agent did today?
Run arcroom init on both machines with the same ArcRoom key and project. The answers, steps and decisions saved during the day live in the ArcRoom API, not on one machine, so the evening agent reads them before it works and reuses them. The last-run note follows the project too, so the desktop starts with where the laptop session stopped.
Also asked as: The agent on my second computer knows nothing about today's work. How do I fix that? · How do I sync AI agent memory between a laptop and a desktop? · Can an AI agent on one machine see decisions saved on another?
Memory lives in the ArcRoom API, not on one machine.
- Decision
[desktop] Saved today on laptop: 4 answers, 1 decision.
Step by step
- Memory lives in the ArcRoom API, not on one machine.
- Both machines run arcroom init with the same key and project.
- The evening session gets the day's saved answers and decisions.
What the agent sees
[desktop] Saved today on laptop: 4 answers, 1 decision.Compared with today's tools
Commit a NOTES.md and push
End the day by having the agent write NOTES.md, commit and push it, and read it at home.
- Setup
- 0 min
- Ongoing
- ~5 min per switch
ArcRoom
Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.
- Setup
- 2 min per machine (npx arcroom init)
- Ongoing
- None
Close when you remember; ArcRoom carries answers, decisions and the last-run note without a commit.
Where it stops
Saved answers, decisions and the last-run note travel; the raw session and its file contents don't.
Try it in your project
Works with the agents your team already uses, and any agent over MCP or the HTTP API.