YesPaPa

YesPaPa is the human-in-the-loop safety layer for AI coding agents. When an LLM agent or script tries to run a destructive command — like rm -rf, git push --force, or DROP TABLE - YesPaPa intercepts it and sends an approval request straight to your phone.

How it works:
- Install the YesPaPa CLI on your machine
- Scan the pairing QR code with this app (or use offline mode with totp)
- Dangerous commands are intercepted and held until you approve

Approve with confidence:
- One-tap approve or deny from a push notification
- See the full command, AI crafted justification, and context before deciding
- Approve with a message to guide the agent's next steps
- Set Auto-Bypass (grace periods) for trusted command patterns

Built on real security:
- TOTP-based authentication (RFC 6238) - works offline too
- Zero-trust architecture - approvals are cryptographically verified
- Tamper-resistant - agents cannot bypass or uninstall their own guardrails(read the docs.yespapa.io for advanced use)
- Falls back to manual TOTP codes if the server is unreachable(self hosting included!)

Perfect for:
- Developers using OpenClaw, Claude Code, Cursor, Copilot workflows, or other AI coding assistants
- DevOps engineers running automated scripts on production systems
- Teams that need human approval gates for destructive operations in CI/CD pipelines
- Anyone who wants a safety net before irreversible commands execute

YesPaPa works with any shell on macOS and Linux. The CLI is open-source at github.com/yespapa-cli.
Docs available: https://docs.yespapa.io