# SafeAgent — Execution Guard for AI Agents
# https://safeagent-production.up.railway.app

## What SafeAgent does
SafeAgent prevents AI agents from executing the same action twice when they retry.
Before any irreversible action fires — payment, trade, email — claim a key.
First call returns PROCEED. Every retry returns SKIP. Atomic, durable, survives process restarts.

## API
POST /claim — gate an action (x402 payment required, $0.001 USDC)
POST /claim/test — free test endpoint, 10 calls per IP
GET /audit — full claim history
GET /health — health check

## State machine
PENDING → COMMITTED → SKIP

## Integration
pip install safeagent-exec-guard
npm install n8n-nodes-safeagent

## Dashboard
https://safeagent-dashboard-2.vercel.app

## EU AI Act Art. 12
SafeAgent provides tamper-evident execution logs for AI agent compliance.