Documentation
Welcome to the AgentX Help Center. Learn how to deploy, manage, and scale autonomous AI agents with our comprehensive guides and architecture deep dives.
Getting Started
Install the CLI and configure your environment variables.
Agents Guide
Explore the capabilities of our 7 specialized autonomous agents.
Elite Prompting
Master the art of mission-based prompting for better results.
Billing & Credits
Understand our credit system and PKR payment protocols.
API Quickstart
1. Get Your API Key
All AgentX API requests require authentication via an API key header. During the Alpha phase, use the public test key below. Production keys will be available in your dashboard after launch.
# Your API Key for Alpha Testing
export AOS_API_KEY="AOS-ALPHA-2026"
# Required Hardware ID header for all requests
export AOS_HW_ID="AOS-V2-STABLE-LOCAL"2. Authenticate a Request
Every request to the AgentX backend must include both the API key and hardware ID in the request headers. The server validates these before processing any operation.
curl -X GET https://api.agentx.io/api/chaos/stats \
-H "X-API-Key: AOS-ALPHA-2026" \
-H "X-Hardware-ID: AOS-V2-STABLE-LOCAL"3. Trigger Your First Chaos Vector
Send a controlled SQL injection payload to test your system's autonomous healing capabilities. The guardian will detect, isolate, and repair the fault automatically.
curl -X POST https://api.agentx.io/api/chaos/attack \
-H "Content-Type: application/json" \
-H "X-API-Key: AOS-ALPHA-2026" \
-H "X-Hardware-ID: AOS-V2-STABLE-LOCAL" \
-d '{"attack_type":"sql_injection","target_url":"https://your-app.com/api/health"}'sql_injection vector is available on all tiers including Free. The server_crash vector requires Enterprise or Founder tier access.4. Network Configuration
For local development, configure your environment to proxy API requests through the Next.js route handlers. This ensures authentication is handled server-side without exposing your API key to the client.
# .env.local
NEXT_PUBLIC_API_URL=http://localhost:8000
NEXT_PUBLIC_SUPABASE_URL=https://your-project.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=your-anon-keyWhy AgentX is Essential
The Core Problem
As AI agents, autonomous LLM chains, and multi-vendor AI microservices scale, they become highly susceptible to runtime drift, alignment failures, unexpected crashes, and external vector exploits. Traditional DevOps firewalls and monitoring tools cannot observe semantic loops, detect prompt injection patterns, or monitor hardware-bound agent states. This creates a dangerous blind spot where production systems can fail silently, inject corrupted outputs, or expose sensitive data without triggering any alert.
The Solution
AgentX serves as the world's first autonomous immune system layer for agentic systems. It acts as an orchestrator that wraps around running agent environments to monitor, simulate chaos, and self-heal anomalies in real time — before they affect production traffic. By continuously polling health status, memory vectors, and processing telemetry, AgentX provides the observability and resilience layer that traditional infrastructure tools cannot deliver.
Who It's For
AI Engineers & DevOps Leads
Eliminates the guesswork in deploying unpredictable agent pipelines. Get real-time visibility into every running agent's health, memory state, and execution metrics.
Security Architects
Full visibility into prompt injection risks, data leakage paths, and local model safety. Chaos vectors let you stress-test defenses before attackers do.
SaaS Enterprise Founders
Maximum uptime assurance. Shields commercial user workflows from system failures with autonomous detection and repair in under 12 milliseconds.
The Deep-Dive Functionality Blueprint
Vanguard Dashboard
The centralized real-time telemetry matrix of AgentX. The Vanguard Dashboard provides a unified view of all active agent loops, incoming transaction payloads, memory bandwidth utilization, and fleet-wide health indicators. It serves as the primary operational interface for both monitoring and manual intervention.
- Real-time fleet map with global node tracking
- Live agent loop monitoring with execution timelines
- Transaction payload inspection and replay
- Manual/Auto hardening toggle for instant lock-down
- WebSocket-powered live updates (no polling)
- Integrated chaos engine control panel
Neural Kernel Integration
The low-level monitoring layer that bridges raw hardware and model state to the dashboard UI. The Neural Kernel tracks underlying agent execution metrics, memory vectors, and processes hardware verification locks to prevent unauthorized instruction changes. It operates at the instruction level, intercepting agent state transitions before they reach the orchestration layer.
- Hardware Verification Lock enforcement
- Memory vector tracking and anomaly detection
- Execution metrics pipeline with sub-ms resolution
- Unauthorized instruction change prevention
- Real-time state transition interception
- Bridge layer between hardware and dashboard UI
Honeypot Traps & Deception Gating
An automated, isolated sandbox designed to capture and analyze malicious payloads targeting your agent infrastructure. The honeypot system feeds simulated vectors to attackers attempting to breach the agent system, tracking their signatures, techniques, and patterns — then automatically blocking their IP and API keys before they can touch actual production databases or agent states.
- Malicious payload capture and isolation
- Attacker signature analysis and fingerprinting
- Automatic IP and API key blocking
- Isolated sandbox environment (zero production impact)
- Deception vector injection to mislead attackers
- Comprehensive attack timeline reconstruction
Cyber Security Vector Simulator (ChaosPanel)
The controlled environment stress-tester. The ChaosPanel allows authorized administrators to trigger SQL injection payload arrays (20 vectors) or server crash fault bombs (6 rapid-fire faults) to evaluate how fast the autonomous guardian system detects, isolates, and repairs the microservice. Every attack is logged with detection latency and repair metrics for compliance reporting.
# Available attack vectors
Vectors:
sql_injection → 20 payload scan [Free+]
server_crash → 6 rapid-fire faults [Enterprise+]
# Response format
{
"status": "DISPATCHED",
"type": "sql_injection",
"tier": "enterprise",
"detection_ms": 11,
"repair_ms": 340
}- SQL Injection with 20 unique payload vectors
- Server Crash Bombs with 6 rapid-fire fault injections
- Autonomous detection with average <12ms latency
- Self-heal verification with full repair metrics
- Tier-based access control (Free / Enterprise / Founder)
- Real-time attack dispatch log with status tracking
Display Your AgentX Shield
Copy-paste our shield badge into your GitHub README to proudly show that your AI infrastructure runs under AgentX autonomous guardrails.
GitHub Badge Generator
Show your AgentX protection status
[](https://agentx.vercel.app)
Was this helpful?