Every day
Workspaces
Every folder you open with Lampson becomes a workspace: its own sessions, memory, scheduled tasks, plugins and servers. You can have several open at once and switch between them without stopping anything.
cd ~/code/shop && lampson # registers ~/code/shop (first time) and opens the terminal on it
cd ~/code/blog && lampson --web # a second workspace, in the browser
The hub§
http://127.0.0.1:8080 is the hub: the workspaces screen. Open one, create one (a folder picker on your desktop; a server-side browser on a VPS), turn them on and off, set how long an idle one stays alive.
Each workspace runs in its own process behind the hub; when the hub restarts by itself (after you add or remove a workspace) they keep running. Every 15 seconds the hub starts what should be running and stops what has been idle (⚙ → hours of inactivity; 0 = never; a workspace with scheduled tasks stays up).
lampson --hub start|stop|status|logs|restart
lampson --install # keep the hub alive across logins (Windows: Scheduled Task · Linux: systemd --user)
Turning Lampson off or restarting it§
Turning a workspace off (on its card in the hub) stops that project's process; the hub stays on at localhost:8080 so you can turn the others on. For Lampson as a whole:
lampson --hub stop # the hub, every workspace and the HTTPS edge (--domain)
lampson --hub start # back on: http://127.0.0.1:8080
lampson --hub restart # stop + start, for example after an update
lampson --hub status # is the hub up, which workspaces are running
lampson --hub logs # the last lines of the hub's log
lampson --uninstall # undo --install: the hub no longer starts at login
The same from the browser: ⚙ → Apagar o reiniciar. A restart comes back on its own and the page reloads; turning it off works only from the PC itself, because from your phone there would be no way to turn it back on.
stop and restart cut the terminals, the processes the agent started and any turn in progress; sessions are kept. If you used lampson --install, a stopped hub starts again at your next login.
Where things live§
Nothing is written into your project. Lampson keeps its own state in ~/lampson:
| path | what |
|---|---|
~/lampson/.lampson/config.json | provider, model, keys, preferences |
~/lampson/.lampson/ws/<name>/ | one folder per workspace: sessions, traces, schedules, process logs |
~/lampson/memory/<name>/ | the agent's notes about that project |
~/lampson/plugins/ | global plugins |
Inside your project, only what you choose to commit: .lampson/plugins/, .lampson/mcp.json, .lampson/lsp.json, skills/, AGENTS.md.
Why a process per project§
A tool's file access in Synsema is a literal folder name declared in the code. A workspace is therefore a small folder with a link named workspace that points at your project; its process only ever sees that folder. That is what makes "it can only touch this project" true rather than hopeful.