Skip to content
TakoAPI
Back to skills

0g-compute

by in-liberty420

Coding Agents & IDEs
531Downloads reported by clawskills.sh
0Views
0 likesSign in to like this skill
—Not yet rated
Your rating

Brief

Use cheap, TEE-verified AI models from the 0G Compute Network as OpenClaw providers.

Description

What This Skill Does

Connects OpenClaw to the 0G Compute Network, a decentralized AI inference marketplace where providers run models inside TEE-verified secure enclaves. Handles wallet management, provider discovery, integrity verification, and OpenClaw configuration. Supports DeepSeek, GLM-5, Qwen, and other models listed on the 0G marketplace.

TEE hardware attestation (Intel TDX) lets you confirm the exact model and code running on a provider before trusting it, which standard API marketplaces don't offer.

When to Use It

  • Comparing 0G inference costs against OpenRouter for the same model
  • Verifying a new provider's TEE attestation before routing traffic to it
  • Funding and monitoring per-provider sub-account balances
  • Adding a verified 0G provider as a custom OpenClaw model endpoint
  • Checking wallet status and current network configuration

Example Workflow

INPUT

User asks: set up a verified 0G provider running DeepSeek as an OpenClaw model

AGENT
  1. 1

    Run 0g-compute-cli inference list-providers to find providers offering DeepSeek

  2. 2

    Check health and TeeML support, pick a candidate provider address

  3. 3

    Run 0g-compute-cli inference verify --provider <address> to confirm TEE attestation

  4. 4

    Run 0g-compute-cli inference get-secret --provider <address> to obtain an API key

  5. 5

    Add the provider entry to openclaw.json with baseUrl, apiKey, and model id, then register an alias in agents.defaults.models

OUTPUT

DeepSeek model available in OpenClaw via the verified 0G provider, billed on-chain

Requirements

  • 0G wallet funded with 0G tokens
  • 0g-compute-cli installed via npm (@0glabs/0g-serving-broker)

Installation

clawhub install in-liberty420/0g-compute

Run the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.

Links