Agent Cloud Overview

Agent Cloud

Agent Cloud is AINative's managed infrastructure for deploying and orchestrating autonomous AI agents. It provides everything agents need to run in production: identity, networking, deployment, and observability.

Architecture

┌─────────────────────────────────────────────────┐
│                 Agent Cloud                       │
│                                                   │
│  ┌──────────┐  ┌──────────┐  ┌───────────────┐  │
│  │ Registry  │  │  Auth    │  │ Deployments   │  │
│  │ & A2A     │  │ OAuth2.1 │  │ Kubernetes    │  │
│  │ Discovery │  │ Tokens   │  │ Auto-scale    │  │
│  └─────┬─────┘  └─────┬────┘  └──────┬────────┘  │
│        │              │              │            │
│  ┌─────┴──────────────┴──────────────┴─────────┐ │
│  │          A2A Networking Layer                 │ │
│  │   Direct messaging · Capability routing      │ │
│  └──────────────────────────────────────────────┘ │
│                                                   │
│  ┌──────────────────────────────────────────────┐ │
│  │          Observability                        │ │
│  │   Traces · Metrics · Cost tracking            │ │
│  └──────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────┘

Modules

Module	Base Path	Purpose
Inference	/api/v1/public/chat, /deployments	Chat completions, dedicated GPU, embeddings, rerank, audio
Dedicated Deployments	/api/v1/public/deployments	Single-tenant H100/MI300X, SLA, auto-scaling, BYOM
Registry	/cloud/agents	Agent registration, discovery, A2A cards
Auth	/cloud/auth	OAuth 2.1 identity for agents
A2A Networking	/cloud/a2a	Agent-to-agent messaging and routing
Task Dispatch	/public/agents/tasks	FIFO task queue with atomic claiming
Deployments	/cloud/deployments	Container deployment and scaling
Observability	/cloud/observability	Traces, metrics, cost tracking

Quick Start

1. Register an Agent

import requests

TOKEN = "your-api-key"
BASE = "https://api.ainative.studio/api/v1/cloud"
HEADERS = {"Authorization": f"Bearer {TOKEN}", "Content-Type": "application/json"}

## Register with capabilities
response = requests.post(f"{BASE}/agents/register", headers=HEADERS, json={
    "name": "my-research-agent",
    "capabilities": ["web_search", "summarization", "analysis"],
    "protocols": ["a2a/1.0"],
    "is_public": True,
    "agent_card": {
        "display_name": "Research Agent",
        "description": "Searches the web and synthesizes findings",
        "version": "1.0.0",
    },
})

agent_id = response.json()["agent_id"]

2. Create OAuth Credentials

## Create OAuth 2.1 client for machine-to-machine auth
client = requests.post(f"{BASE}/auth/clients", headers=HEADERS, json={
    "agent_registration_id": agent_id,
    "scopes": ["agent:read", "agent:write", "memory:read"],
    "token_ttl_seconds": 3600,
})

client_id = client.json()["client_id"]
client_secret = client.json()["client_secret"]  # shown only once

3. Deploy to Cloud

deployment = requests.post(f"{BASE}/deployments", headers=HEADERS, json={
    "agent_registration_id": agent_id,
    "image_uri": "ghcr.io/my-org/research-agent:latest",
    "resource_plan": "standard",
    "runtime_config": {
        "env": {"LOG_LEVEL": "info", "ZERODB_PROJECT": "my-project"}
    },
})

4. Send Agent-to-Agent Messages

## Direct message to another agent
response = requests.post(
    f"{BASE}/a2a/{target_agent_id}/message",
    headers=HEADERS,
    json={
        "message_type": "task",
        "content": {"query": "Summarize recent AI research papers"},
    },
)

## Or route by capability (the platform finds the best agent)
response = requests.post(f"{BASE}/a2a/route", headers=HEADERS, json={
    "capability": "summarization",
    "content": {"text": "Long document to summarize..."},
})

Resource Plans

Plan	vCPU	Memory	GPU	Price
basic	0.5	512 MB	—	Free tier
standard	1	1 GB	—	Usage-based
gpu	2	4 GB	T4	Usage-based

Next Steps

• Inference — Run models, dedicated GPUs, embeddings, rerank, audio
• Dedicated Deployments — Single-tenant H100/MI300X with SLA and BYOM
• Registry API — Register and discover agents
• Auth API — OAuth 2.1 for agents
• Task Dispatch — FIFO task queue with atomic claiming
• Agent SDK — TypeScript client for all Agent Cloud APIs