---
title: "Platform: Runtime Enforcement"
description: "Vero enforces your policies as an agent acts. Read-only MCP access, no secrets for coding agents, and malicious skills and prompt injection stopped."
canonical: https://www.vero.security/platform/runtime-enforcement
last_updated: 2026-11-18T13:15:20Z
---

# Platform: Runtime Enforcement

*Platform · 04 · Runtime Enforcement*

## Stop the risky action. Enable the tool.

Vero enforces your policies at the moment an agent acts, and checks how every agent is set up. An agent can read through an MCP server without being able to write. Coding agents never see your secrets. Malicious skills and prompt injection are stopped, at rest and at runtime. Everyone keeps the AI tools they rely on.

[Book a demo](https://www.vero.security/demo) · [See the platform](https://www.vero.security/platform)

*Today*

A coding agent has a production token in its context. The block list names the tool, but not the action it is about to take.

Without it, a policy is a document. Agents act every second.

*What Vero does*

## You set the goals. Vero does the work.

### 01: At the moment of action
Each tool call is checked against your policies before it runs, not in a weekly report.

### 02: In the configuration too
Settings, permissions and autonomy modes are held to your policies, on every agent.

### 03: The right response
Monitor, Notify, Warn or Block, chosen per policy, moving from monitor to block in steps.

### 04: Where agents run
Agent hooks, LLM and MCP gateways, the network, browser and endpoint, and identity. Nothing new to install.

*How it works*

### Step 1: Monitor
Watch the policy against real activity first.

### Step 2: Warn
Tell people before anything is stopped.

### Step 3: Block
Enforce, once you know no one’s work breaks.

*Proof*

## A Telegram plugin slipped past “official marketplace only.” Vero caught it.

A customer allowed plugins from the official marketplace only. A Telegram plugin still got through the setting. It gave agents a direct line to an outside messaging app: a path for company data to leave. Vero caught it.

*In the solutions*

## *Runtime Enforcement* at work

Each solution is one AI security problem Vero handles for you. These three lean on Runtime Enforcement the most.

### Data Leakage & Privacy
Company data and credentials kept away from model training, personal accounts and outside endpoints.
[See the solution](https://www.vero.security/solutions/data-leakage-and-privacy)

### Agentic Supply Chain
Every MCP server, skill, plugin and package checked before it runs.
[See the solution](https://www.vero.security/solutions/supply-chain)

### Agent Integrity
Prompt injection, tampered instructions and hijacked agents, stopped.
[See the solution](https://www.vero.security/solutions/agent-integrity)

## *More from* the platform

### Oversight
[Read](https://www.vero.security/platform/oversight)

### AI Security Expertise
[Read](https://www.vero.security/platform/ai-security-expertise)

### Security Goals
[Read](https://www.vero.security/platform/security-goals)

### People Workflows
[Read](https://www.vero.security/platform/people-workflows)

*Get started*

## You set the goals. Vero does the work.

[Book a demo](https://www.vero.security/demo)
