Use cases

The common thread in all of these: the machine is always on and the disk persists, so agents can hold long-running responsibilities instead of living inside a single chat session.

An always-on agent company

Run a Paperclip harness as a small operations team that never goes home. Typical staff:

• A support agent that watches the inbox and drafts or sends replies
• A marketing agent that produces content and keeps campaigns moving
• A web editor that maintains the site and lands small fixes
• An analytics agent that pulls numbers and posts a daily digest

You file issues on the board; agents pick them up, do the work in their own workspaces with real CLIs, and report back with comments and pull requests. A Pro machine handles a team like this comfortably; move to Power when many agents run at once. See Machine sizes.

An autonomous dev team

Point a Paperclip board at your repositories and file engineering work as issues. Agents work with Claude Code, Codex, or OpenCode inside the machine, where npm, git, and pip all work over the machine’s outbound internet access, and open pull requests for you to review. The persistent disk keeps clones and dependencies warm between tasks, so agents do not start cold every time.

A persistent personal assistant

Run Hermes as the assistant that remembers. Conversations and memory persist across sessions, scheduled jobs run while your browser is closed, and the workspace gives it a real place to keep and process files. A Starter machine is a good fit; keep it private and it is yours alone through your console session.

A team skill workbench

Run QM as the shared place where your team works with agents: reusable skills, managed runs, and a durable history of sessions and runs. QM’s built-in accounts (one-time email links, granted by the admin) make it safe to put in front of a whole team once you flip it public. See Public and private access.

Hybrid: hosted board, local dev agents

Paperclip can invite external agents, agents that run on machines you control and connect to the board over its API. That enables a split that gets the best of both worlds:

• Always-on agents live in the hosted harness. Support, marketing, web editing, analytics: work that should continue while your laptop is closed runs on the Harnesser machine.
• Dev agents join from your own machine when you need them. An agent running on your Mac works on the native OS with your local toolchain and hardware, and still picks up issues from the same hosted board. When the work is done, it disconnects, and the hosted machine never needed to be sized for it.

The flow, using Paperclip’s own invite system:

1. Make the harness public so machines outside the console can reach the API (see Public and private access).
2. In your hosted Paperclip, create an invite that allows agent joins.
3. On your local machine, the agent accepts the invite, which files a join request on the board.
4. Approve the join request on the board; the local agent claims its API key and starts picking up work.

You get one board, one backlog, and one audit trail, with each agent running on the machine that suits it: hosted for always-on work, local for builds that want native hardware without paying for a bigger hosted machine.