you need to try Paperclip RIGHT NOW!

Paperclip functioned as a task-based coordination layer for separate Claude Code, Hermes, Codex, and local-model agents [09:17]. The network diagnosis remained an unverified working conclusion rather than proof that the toilet flush caused the disconnects [22:52].

Key points

• The agent artifacts said the NAS never went down, the router was not down, and no machine settings changed [18:11]. The initial report could not name the toilet as the trigger, although it noted a shared physical disturbance could explain the observed pattern [18:24]. A later report tied a shared physical event to four specific fiber runs dropping within 2 seconds, 36 times, which pushed the diagnosis toward that switch [18:45].

• Ports 3, 5, 13, and 19 lost synchronization 56 times per day, and port 5 logged 20,052 link-down events since boot versus 123 for a comparable port [20:17]. Occupancy-linked disturbances knocked two to four links down in the same second seven times each working day, which helped isolate the affected ports rather than the whole switch [20:40]. The fix was replacing those third-party SFP modules with MikroTik-branded modules from a different vendor/batch [21:23].

• Paperclip is not itself the model harness; it coordinates existing harnesses such as Claude Code, Hermes, Codex, and local models through assigned tasks [06:54]. The CEO agent delegated work to remote employees, and blocked or finished work appeared as tasks, artifacts, and decisions instead of a single free-form chat [09:17]. Scheduled routines could trigger recurring checks [17:25]. The organization could be exported or imported as a reusable packet [25:30].

Step by step

1. Install Paperclip on Mac, Linux, WSL2, or a dedicated Ubuntu VM/VPS by downloading and running the installer, then open port 3100 to onboard the company [06:44].
2. Create a CEO agent on a chosen harness, such as Claude Code, and hire remote employee agents on other harnesses, approving each hire in the interface [07:12].
3. Create a project and root task, assign it to the CEO, and let subtasks be delegated; review results through artifacts, decisions, and approval points [11:33].
4. Add a routine with an explicit trigger, such as daily at 10:00, assign a responsible agent, and require a decision when storage or health thresholds are breached [17:25].
5. Export the organization from the org screen to transfer agents, routines, tasks, and attachments to another Paperclip instance [25:30].