Paperclip explained: an operating layer for teams of AI agents
Deploying Paperclip: local trust and authenticated access
Select the security mode before binding the dashboard to a network.
What you will learn
- Separate identity from reachability
- Move data and secrets deliberately
- Make deployment checks observable
Before you start
- Node.js and agent runtime basics
- A test company with non-sensitive tasks
Turn the source review into a bounded experiment with stop conditions.
Key takeaways
- Loopback and authentication solve different problems.
- Board-claim links deserve secret handling.
- Database recovery must be tested on the chosen deployment.
Separate identity from reachability
The deployment guide distinguishes local_trusted from authenticated. The first binds to loopback without a login; the second can run behind a private network or at a public URL with stricter checks. A bind address alone does not establish an authorization policy.
For a team, document the login method, base URL and reverse-proxy trust boundary. Test that an unauthenticated browser cannot reach board data before inviting people.
Move data and secrets deliberately
The README describes an embedded PostgreSQL path for local use and an external database for production. Record the database location, backup process, secret source and agent workspace path before an upgrade.
When switching modes, the deployment guide describes a one-time board-claim URL. Treat that link as a credential, restrict its audience and verify ownership after the claim.
Make deployment checks observable
Check startup, login, company membership, one harmless heartbeat and a restore from backup. The README also recommends doctor checks for public mode; run them on the actual host.
No deployment was performed for this review. The guide explains supported paths, but it cannot certify your proxy, database or agent sandbox.
Decision guide
| Criterion | Option A | Option B |
|---|---|---|
| Best when | You need predictable behavior and easy auditing | You need adaptive optimization and have reliable telemetry |
| Main risk | May leave performance on the table | Can become difficult to explain or debug |
Implementation steps
- 1
Choose local_trusted only for a single-user loopback trial.
- 2
For sharing, configure authenticated mode and verify the public or private URL.
- 3
Rehearse board claim, backup restore and one harmless run.
Copy-ready example
pnpm paperclipai configure --section server
# select authenticated/private or authenticated/publicFrequently asked questions
Can I place local_trusted behind a public reverse proxy?
That would expose a mode with no login; use authenticated public mode instead.
Is external Postgres mandatory for the first test?
The README describes an embedded option for local work; production requirements decide later.
Sources
- Paperclip / README.mdSource checked 2026-09-26
- Paperclip / docs/deploy/deployment-modes.mdSource checked 2026-09-26