Nightly checks
Let a remote Claw run at night and wake you only when needed
Put routine diagnostics and log summarization on a persistent host, then pull a human back in only for real decisions.
Starting condition
One always-on Linux Claw is already connected with Trigger and message center configured.
- A scheduled Trigger launches the Agent overnight.
- The Agent aggregates evidence before deciding whether a human needs to be paged.
- You continue from desktop or mobile once a real decision is needed.
Automation handles the routine; attention goes only to the exceptions.
Split host duties
Patch from one machine, validate from another
Keep code changes and verification on separate hosts instead of constantly switching shells and contexts yourself.
Starting condition
Your development machine and an internal verification machine are both connected as Claws.
- Use the local Claw to edit code and prepare the patch.
- Switch execution to the internal Claw for tests or service verification.
- Keep the whole flow inside one conversation timeline.
Context stays stable while execution environments remain flexible.
Subagent pair
Let one Agent design while another writes the patch
Keep the main thread for control while specialized subagents move implementation forward in parallel.
Starting condition
At least one subagent workflow or conversation agent is available.
- Split the work into design, implementation and review roles.
- Each subagent returns focused output.
- Merge the result back into the main thread for the final decision.
Large tasks become flatter and easier to govern.
External events
Wake an Agent directly from another system
Webhook or SDK triggers let CI, internal services or operational tools initiate work automatically.
Starting condition
A webhook-based Trigger or SDK integration is configured.
- Another system pushes an event into Admin.
- Trigger selects the right Agent and context.
- The result returns through message center or a callback.
Agents become event-driven execution units instead of only manual chats.
Memory continuity
Resume a project at night without retelling the whole story
clue / dream and layered memory indexes bring important constraints into the next session.
Starting condition
Long-term memory capabilities are enabled and have already collected useful project context.
- Important details from a daytime session become memory material.
- A new evening session recalls the relevant constraints first.
- You only add what changed today.
Long-running work becomes genuinely continuous across conversations.
Mobile follow-up
Catch a confirmation on mobile and keep work moving
When an Agent needs a yes/no decision, a path or a risk choice, you should not have to go back to your desk first.
Starting condition
Mobile client is logged in and push permissions are configured.
- A desktop task hits a decision point.
- Message center and push deliver the request to mobile.
- You answer from mobile and the desktop-side task continues.
Work stops less often because human intervention becomes lighter.
