Security Integration for AI Agents

Three integration paths: Python SDK for integrators, CLI for shell agents, and REST API for other languages. Choose the approach that fits your architecture.

Python SDK

Native Python SDK for integrators. Import and use directly in your code.

  • Direct Python import
  • Full scanning + Shield
  • Type hints included
  • Auto-escalation logic
  • Best for Python projects
import agentguard360

AgentGuard360 CLI

Command-line interface with agent mode for JSON output. Best for agents with shell access.

  • JSON output mode
  • Shell-accessible agents
  • Local Shield scans (free)
  • No API key management
  • Best for shell workflows
agentguard360 --agent

Direct API

RESTful HTTP API for any language. Use when SDK/CLI aren't available.

  • Language agnostic
  • RESTful JSON API
  • Pay-per-scan (x402)
  • OpenAPI spec available
  • Best for JS, Go, etc.
POST /v1/guard

Python SDK

Native Python SDK for integrators. Install once, import anywhere.

Content Scanning

PYTHON

Scan content with auto-escalation to premium if threats detected.

import agentguard360 result = agentguard360.scan("Check this content for threats") print(f"Verdict: {result.verdict}, Threats: {result.has_threats}")
Verdict: clean, Threats: False

URL Safety Check

PYTHON

Check URL reputation before fetching content.

url_result = agentguard360.check_url("https://example.com/api") print(f"Safe: {url_result.is_safe}, Risk: {url_result.risk_level}")
Safe: True, Risk: low

Shield Scan (Device Hardening)

PYTHON

Scan local environment for security vulnerabilities. Runs locally (free).

shield = agentguard360.shield_scan(mode="rapid") print(f"Score: {shield.score}/100 ({shield.grade})") print(f"Critical: {shield.critical_count}, Warnings: {shield.warning_count}")
Score: 85/100 (B) Critical: 0, Warnings: 2

Detect AI Agents

PYTHON

Find installed AI agents and their configurations.

agents = agentguard360.detect_agents() for agent in agents: print(f"{agent.name}: active={agent.is_active}")
Claude Code: active=True Cursor: active=False

AgentGuard360 CLI

All commands in agent mode return structured JSON for easy parsing. No interactive prompts.

Discover Tools (LLM-Native Schemas)

ANTHROPIC / OPENAI

Returns tool schemas ready for direct registration with Claude, GPT, or any tool-calling model.

agentguard360 agent discover # Anthropic format agentguard360 agent discover -f openai # OpenAI format
{ "version": "1.0.0-beta", "tools": [ { "name": "scan", "description": "Scan content for security threats...", "input_schema": { "type": "object", "properties": { "content": {"type": "string", "description": "Text to scan"} }, "required": ["content"] } }, // ... 15 more tools ], "tool_count": 16 }

Call Any Tool

JSON

Execute any of the 16 security tools directly. All return structured JSON.

agentguard360 agent call shield_scan agentguard360 agent call scan --args '{"content": "text to check"}' agentguard360 agent call check_url --args '{"url": "https://example.com"}'
{ "success": true, "output": { "score": 85, "grade": "B", "critical_count": 0, "warning_count": 2, "issues": [...] }, "error": null }

List Available Tools

JSON

Quick list of all 16 available tool names.

agentguard360 agent tools
{ "tools": [ "scan", "risk_assess", "radar_scan", "check_url", "check_urls", "scan_document", "shield_scan", "detect_agents", "detect_mcp_servers", "get_risk_credits", "purchase_risk_credits", "get_wallet_balance", "is_wallet_configured", "get_quote", "get_pricing", "get_status" ], "count": 16 }

Why LLM-Native Discovery?

FIRST OF ITS KIND

Most Python packages require agents to parse docs or scrape CLI output. AgentGuard360 returns schemas in the exact format LLMs expect—ready to register as callable tools.

# Agents don't parse documentation—they discover and call agentguard360 agent discover # Get schemas agentguard360 agent call scan --args '{"content": "..."}' # Execute

What Agents Can Scan

Security scanning across the content types AI agents commonly process.

Conversations

Chat messages, email threads, support tickets. Detect social engineering and manipulation attempts.

Documents

PDF, DOCX, code files. Hidden instructions embedded in formatting, metadata, or invisible text.

URLs & Web Content

Fetch and scan web pages before ingestion. Catch threats at the source.

API Responses

Third-party API data, webhook payloads. Validate before processing.

Activity Intelligence

ML-powered behavioral analysis for runtime activity monitoring.

Anomaly Detection

Flags unusual patterns—sudo spikes, unexpected uploads, abnormal API calls.

Risk Trajectory

Predicts where your risk score is heading based on recent activity trends.

Breach Probability

Estimates likelihood of security incident in the next 7 days.

Community Comparison

See how your agent ranks vs. anonymized community baselines.

Skill.md for Context Injection

For agents using direct API integration, paste these instructions into the agent's context or system prompt.

Loading instructions...

Machine-Readable API Specification

Agents can fetch the complete API specification directly:

GET https://aisecurityguard.io/v1/skill

Returns JSON with all endpoints, parameters, threat categories, and pricing. OpenAPI spec also available at /openapi.json.

Get Started Now

Install AgentGuard360 and run your first scan. No account required for Shield scans.

pip install agentguard360