How Agent Action Firewall Works
A simple, powerful architecture that sits between your AI agents and the actions they take. Every action is evaluated, controlled, and logged.
The Architecture
Agent Action Firewall intercepts every action before it executes, giving you complete control.
AI Agent
Your AI agent wants to perform an action (API call, file write, etc.)
Agent Action Firewall
Evaluates action against your policies in real-time
Decision
Allow, deny, or require human approval
Three Pillars of Control
Policy Evaluation
Define what your agents can and cannot do using flexible policies. Actions are evaluated in milliseconds against your rules.
- Action type filtering
- Resource-based rules
- Time-based restrictions
Approval Workflows
Risky actions pause and wait for human approval. Get notified via Slack, email, or other channels. Approve or deny with a single click.
- Real-time notifications
- Configurable timeouts
- Approval delegation
Audit Trail
Every action, decision, and approval is logged with cryptographic integrity. Search, filter, and export for compliance.
- Tamper-evident logs
- Exportable reports
- Full-text search
Simple Integration
Add Agent Action Firewall to your AI agent with just a few lines of code. Our SDK handles all the complexity.
Install the SDK
Add our TypeScript or Python SDK to your project
Wrap Your Actions
Submit actions through the firewall before executing
Define Policies
Set up your security policies in the dashboard
import { AgentFirewall } from '@agent-firewall/sdk';
const firewall = new AgentFirewall({
apiKey: process.env.AAF_API_KEY
});
// Before executing any action
const decision = await firewall.check({
type: 'api_call',
target: 'https://api.example.com',
method: 'DELETE',
resource: '/users/123'
});
if (decision.allowed) {
// Execute the action
await executeAction();
}Common Use Cases
Agent Action Firewall protects AI agents across many scenarios
Code Assistants
Prevent AI coding assistants from executing dangerous commands or accessing sensitive files.
Customer Service Bots
Require approval before refunds, account changes, or escalations to prevent costly mistakes.
Data Processing Agents
Control which databases agents can query and what data they can export or modify.
DevOps Automation
Prevent unauthorized deployments, infrastructure changes, or access to production systems.
Research Assistants
Control which external APIs and data sources research agents can access.
Content Generation
Review and approve AI-generated content before it gets published publicly.
Ready to Secure Your AI Agents?
Start protecting your AI agents in minutes. Free tier available.