All Systems Operational

Connect Your Agent in Seconds

No accounts. No API keys. Just your Solana wallet. Connect via MCP and start running inference immediately.

Get Started โ†’

โšก Quick Start

Connect your agent in 3 lines of code

Python
from mcp import ClientSession

# Connect to ASG Agent Cloud MCP server
async with ClientSession("wss://mcp.asgcompute.com") as session:
    
    # Run inference - wallet IS your identity
    result = await session.call_tool(
        "run_inference",
        {
            "model": "deepseek-v3",
            "messages": [{"role": "user", "content": "Hello!"}],
            "payment_wallet": "YOUR_SOLANA_WALLET"
        }
    )
    
    # If 402 Payment Required โ†’ sign USDC transaction
    # Then retry with payment proof
    print(result.content)

๐Ÿ“– See full documentation at /developers

๐Ÿง  Supported Models

Enterprise-grade inference at unbeatable prices

Model Best For Price / 1K Tokens Latency
DeepSeek V3 Best Value Coding, General $0.001 USDC ~50ms
Qwen 3 235B Flagship Complex Reasoning $0.003 USDC ~120ms
Llama 3.3 70B Fast Inference $0.0008 USDC ~30ms
MiniMax M2.1 Agentic Multi-step Tasks $0.005 USDC ~80ms

๐Ÿ’ฐ Pricing

Pay per request in USDC. No minimums, no commitments.

๐Ÿง  Inference

$0.001
per 1K tokens (DeepSeek V3)

๐Ÿ–ฅ๏ธ Dedicated GPU

$4.50
per hour (H200 141GB)

โšก Serverless Modal

Pay-per-sec
Custom Model Hosting

๐Ÿ“š RAG Query

$0.005
per query
โš™๏ธ Code Sandbox
$0.01
per minute (CPU)

Full Pricing Calculator โ†’

๐Ÿ”ง Available MCP Tools

Everything your agent needs via standard MCP protocol

MCP Tools
get_pricing()       # Get current prices for all resources
check_balance()     # Check USDC balance for a wallet
run_inference()     # Execute standard model inference (OpenRouter)
run_modal()         # Execute custom serverless models
provision_gpu()     # Allocate dedicated GPU resources
terminate_compute() # Stop GPU resources
create_rag_index()  # Create a knowledge base
query_rag()         # Query a knowledge base