Raqvio Agent Guardrails documentation

Raqvio Agent Guardrails is a WordPress plugin that decides what AI agents may do on your site. WordPress 6.9 added the Abilities API: plugins (WordPress core, WooCommerce and others) register abilities such as “update a product” or “change an order status”, and AI agents discover and run them over REST or MCP. Raqvio Agent Guardrails sits in front of every ability execution and answers one question: should this agent be allowed to do this, right now?

How a call is decided

1. Identify the caller. Raqvio Agent Guardrails maps the request to an agent you created (by application password or WordPress user), to the built-in Site AI for AI features running inside WordPress, or to an unknown agent. Logged-in humans are not governed unless you turn on “Govern all ability executions”.
2. Decide. The emergency kill switch comes first, then paused agents, then (Pro) policy rules, then the agent’s permission for that ability, then the agent’s default for unlisted abilities, then the risk-based defaults (used for unknown agents and agents set to “Use risk defaults”). The result is allow, deny, or (Pro) require approval.
3. Record. Every governed call (except Raqvio Agent Guardrails’s own read-only informational abilities) is written to the audit log with its agent, channel, decision, status and redacted input.

On WordPress 7.1 and newer the decision is enforced. On 6.9 and 7.0 Raqvio Agent Guardrails runs in audit-only mode: it logs what it would have decided but cannot block anything.

Start here

Install & requirementsVersions, activation, who can manage Raqvio Agent Guardrails.

Quick startGovern your first agent in about ten minutes.

Connect an agentSet up MCP, then connect Claude Desktop, Claude Code or Cursor.

Free vs ProWhat each edition includes, and how to upgrade.

Screens in wp-admin

Raqvio Agent Guardrails adds an Agent Guardrails menu with these screens:

Screen	What it is for	Edition
Dashboard	Kill switch, last 24 hours, abilities by risk, agents, hardening, readiness, recent agent activity	Free
Abilities	Every ability on the site, its risk level and exposure	Free
Agents	Agent profiles and per-ability permissions	Free
Rules	Conditional policy rules, spending budgets and templates, with a test on past activity	Pro
Approvals	Agent actions waiting for a human	Pro
Audit Log	Every governed call, with redacted input	Free (filters, export, integrity check and rollback in Pro)
Readiness	Catalog score for AI shopping agents (WooCommerce only)	Free (fixer and feed in Pro)
Settings	Defaults, retention, notifications, alerts, access, hardening checklist	Free (notifications and alerts in Pro)

The admin bar also shows an Agents item with the kill switch toggle and, with Pro, the number of pending approvals.