0G Compute Network

Interface with the 0G Compute Network — a decentralized AI inference marketplace with TEE-verified model integrity.

Prerequisites

• 0g-compute-cli installed: npm i -g @0glabs/0g-serving-broker
  • Note: npm package name is @0glabs/0g-serving-broker (the old @0glabs/0g-compute-cli package name no longer resolves), but the binary command is still 0g-compute-cli.
• Wallet funded with 0G tokens
• Logged in: avoid passing key on the command line when possible (it can be visible in shell history/process list). Prefer reading from a protected prompt/env:
  • read -s OG_PK; 0g-compute-cli login --private-key "$OG_PK"; unset OG_PK
• Network configured: 0g-compute-cli setup-network

Core Workflows

1. Discover Models

# List all providers with models, prices, verifiability
0g-compute-cli inference list-providers

# Detailed view with health/uptime metrics
0g-compute-cli inference list-providers-detail

# Include providers without valid TEE signer
0g-compute-cli inference list-providers --include-invalid

Filter output by model name, price, health status, and TeeML support (models running in Trusted Execution Environment).

2. Verify Provider Integrity

Always verify before trusting a new provider. TEE verification ensures the model runs in a secure enclave with hardware-attested integrity.

# Full TEE attestation check
0g-compute-cli inference verify --provider <address>

# Download raw attestation data
0g-compute-cli inference download-report --provider <address> --output report.json

The verify command checks:

• TEE signer address matches contract
• Docker Compose hash integrity
• DStack TEE (Intel TDX) attestation

3. Wallet & Balance

# Account overview
0g-compute-cli get-account

# Per-provider sub-account balance
0g-compute-cli get-sub-account --provider <address> --service inference

# Fund operations
0g-compute-cli deposit --amount <0G>                          # To main account
0g-compute-cli transfer-fund --provider <addr> --amount <0G> --service inference  # To sub-account
0g-compute-cli retrieve-fund --service inference              # From sub-accounts
0g-compute-cli refund --amount <0G>                           # Withdraw to wallet

Important: Inference calls fail if sub-account balance is depleted. Monitor balances regularly.

4. Configure OpenClaw Provider

Get API key from a verified provider (interactive — prompts for expiration):

0g-compute-cli inference get-secret --provider <address>

Add to openclaw.json:

"providers": {
  "0g-<model-name>": {
    "baseUrl": "<provider-url>/v1/proxy",
    "apiKey": "<secret>",
    "api": "openai-completions",
    "models": [{ "id": "<model-id>", "name": "<display-name>" }]
  }
}

Register in agents.defaults.models with an alias. Set cost: 0 since billing is on-chain.

See references/openclaw-config.md for complete setup guide.