Policy
AI governance
Paperclip is the AI backend for agent work. AI can draft, classify, summarize, and recommend; it cannot publish, send, submit, pay, or expose sensitive locations without human approval.
Safe autonomous work
- recommendation scoring
- tag suggestions
- duplicate detection
- listing quality scoring
- saved-search alerts from verified data
- internal summaries
Protected actions
- vendor application submission
- external cold email send
- payment submission
- legal/privacy response
- sensitive exact-location publication
- removal decisions
- public affiliation claims
Approval gate
Application drafts, outreach drafts, SEO suggestions, and safety screens are approval-required: yes.
Paperclip webhook actions are authenticated with a shared secret, written to review queues when durable action is needed, and routed through admins for protected changes.