Skip to content

Security

Security Guardrails for AI Agents on Production WordPress

· 2 min read · AI Bridge

Cybersecurity lock icon on digital background
AI Bridge ships with guardrails — but connected agents still have administrator-level power.

Connecting an AI agent to WordPress is not like installing a read-only analytics plugin. A authenticated MCP session can run PHP, write files, and modify the database. AI Bridge is designed for staging and development first — with guardrails that reduce risk when you enable abilities on production.

Threat model: what can go wrong

  • Agent deletes or overwrites critical theme files
  • Runaway PHP exhausts memory or causes fatal errors
  • Credentials leak through chat logs or shared MCP configs
  • Production changes without git review or backup

Understanding these scenarios helps you configure guardrails appropriately.

Built-in AI Bridge protections

Off by default

Installing the plugin exposes nothing. An administrator must explicitly enable AI abilities and confirm the security implications.

Application Password authentication

Every REST and MCP request requires HTTPS and a WordPress Application Password tied to an admin account. There is no anonymous tool access.

PHP sandbox and crash recovery

New PHP files written by agents execute in a sandbox. Fatal errors roll back sandbox state instead of white-screening the entire site.

Protected core paths

wp-config.php, WordPress core files, and the AI Bridge plugin itself cannot be deleted. The plugin refuses to deactivate mid-session.

Kill switch

A secret URL, known only to the site owner, instantly disables all AI abilities from any browser — no wp-admin login required.

Developer reviewing security settings on multiple screens
Rotate Application Passwords regularly and revoke unused credentials after each project.

Recommended practices checklist

  • ☐ Use staging or local installs before production automation
  • ☐ Create dedicated Application Passwords per agent or team member
  • ☐ Store kill-switch URL in your password manager
  • ☐ Review agent file changes in git before deploy
  • ☐ Keep daily backups (Hostinger, UpdraftPlus, or similar)
  • ☐ Test kill switch after enabling abilities
  • ☐ Never paste Application Passwords into public chat logs

Console and security

If you use the AI Bridge Console, credentials are encrypted at rest on your machine. The Console does not replace Application Password auth — it stores and forwards them over HTTPS to WordPress.

Security is a process, not a checkbox. AI Bridge gives you the controls; your workflow determines the outcome.

← Back to blog