Experimental features
Paperclip ships some features behind opt-in flags before they become default behavior. They're real, working features — you can turn them on today — but they're still being evaluated against real usage, so they live on a dedicated Experimental settings page instead of being on for everyone.
This section has one page per experimental feature: why it exists, how to turn it on, and how to use it.
What "experimental" means
An experimental feature in Paperclip:
- Has shipped and works. These aren't stubs — each flag gates a complete surface.
- Is opt-in. Everything stays hidden until you flip the flag, so the UI stays out of your way if you don't use it.
- Comes with no compatibility guarantees. The app puts it plainly when you open the page: "Experimental features may break at any time. These features are opt-in and come with no compatibility guarantees. They may change, break, or be removed without notice. Avoid relying on them for critical or production workflows."
Turning a flag on is not dangerous in the "this will corrupt your data" sense, and flipping one back off is always safe — features are hidden, not deleted, and any data they created is kept. But don't build a workflow that has to stay stable on top of one, and expect to re-read the release notes after an upgrade: experimental features may be renamed, reworked, promoted to core settings, or retired between releases.
Turning a feature on
All experimental flags live in one place, and they're instance-wide (they apply to every company on your Paperclip instance):
- Open Settings → Instance settings → Experimental.
- Find the feature's card and flip its toggle.
Changes take effect immediately — no restart, no migration. Each feature page in this section repeats the exact toggle name to look for.
The features
| Feature | What it adds |
|---|---|
| Environments | Environment management in settings, plus project and agent environment assignment — including custom sandbox images. |
| Isolated Workspaces | Execution-workspace controls: per-run isolated copies of a project so agents don't step on each other. |
| Experimental File Viewer | Task-detail controls for browsing and previewing workspace files relative to a task. |
| External Objects | Detects external URLs in issues and shows live status for referenced pull requests, tickets, and other work objects. |
| Task Plan Decomposition Panel | Shows accepted-plan decomposition history on task detail pages. |
| Task Watchdogs | Watchdog agents that verify stopped task subtrees and restore live paths when work should continue. |
| Cloud Sync | Connect a local instance to a Paperclip Cloud upstream: preview, push, retry, and activation review. |
| Server Info Debug View | A "Server" section in the account drawer with the server restart time and running commit. |
| Auto-Restart Dev Server When Idle | For Paperclip developers: restarts a stale pnpm dev:once boot once all local runs finish. |
| Auto-Create Recovery Tasks | Lets the scheduler create recovery tasks for stalled task dependency chains. |
Where to go next
- Instance settings — the rest of the instance-level settings pages, including where the Experimental page sits.