AgentCDN Developer Documentation

Welcome to the AgentCDN platform documentation. AgentCDN provides a globally distributed runtime that bridges autonomous AI agents, LLM providers, and real-world tools with ultra-low latency.

1. Quickstart Guide

Integrating AgentCDN into your agent framework takes less than 3 minutes. The client library automatically discovers the nearest Anycast edge node and establishes an encrypted connection multiplexer.

Installation

Bash Terminal
# Install via npm
npm install @agentcdn/sdk

# Or install via pip for Python workflows
pip install agentcdn

2. Edge Network Architecture

Traditional content distribution networks are optimized for static caching of immutable assets (images, CSS, video segments). In contrast, AgentCDN implements:

3. Authentication & API Credentials

All HTTP requests to the AgentCDN API must include your project token in the Authorization header:

Authorization: Bearer ak_live_xxxxxxxxxxxxxxxxxxxxxxxx

We strongly recommend storing your API credentials in environment variables and utilizing restricted-scope keys for production worker runtimes.

4. Edge Route API Specification

Dispatch an agent tool call or inference request to the nearest available cluster:

POST /v1/agents/route
POST /v1/agents/route HTTP/2
Host: agentcdn.xyz
Authorization: Bearer ak_live_xxxx
Content-Type: application/json

{
  "agent_id": "agent-lead-01",
  "pop_preference": "anycast",
  "timeout_ms": 2500,
  "payload": {
    "tool": "database_query",
    "query_hash": "sha256_9b8a1c"
  }
}

5. Zero-Trust Security Sandbox

To protect backend systems against prompt injection and malicious autonomous tool execution, AgentCDN incorporates:

  1. Cryptographic attestation for all agent payload hashes.
  2. Strict rate limiting and anomaly scoring at the Edge WAF.
  3. Automated TLS 1.3 encryption with ECC-384 root certificates and Ephemeral Diffie-Hellman key exchanges.