@foundryprotocol/0gkit-cli

The neutral 0g command line — init, doctor, chain, storage, infer, da, attest. Language-agnostic: any stack shells out; --json for scripting. Foundry is a separate, opt-in plugin, never required.

Install

# Recommended: install once, then use the short `0g` binary
npm install -g @foundryprotocol/0gkit-cli
0g doctor

# Or one-off via npx (use the full scoped name — `npx 0g` resolves to an
# unrelated package on npm):
npx @foundryprotocol/0gkit-cli doctor

60-second start (no funds, testnet)

npx @foundryprotocol/0gkit-cli init my-app && cd my-app
npm install
npx @foundryprotocol/0gkit-cli doctor                 # preflight every 0G surface
npx @foundryprotocol/0gkit-cli chain faucet 0xYourAddress   # Galileo → points you at https://faucet.0g.ai

Commands

| Command | What | | ------------------------------ | -------------------------------------------------------- | | 0g init [name] | scaffold a runnable, testnet-default project | | 0g doctor | RPC / signer / storage / DA / faucet checklist | | 0g chain faucet\|balance\|tx | faucet guidance, native balance, await a receipt | | 0g storage put\|get\|exists | upload/download/probe 0G Storage | | 0g infer | chat completion against a 0G compute provider | | 0g da publish\|verify | publish a blob / local integrity check | | 0g attest verify\|report | verify or summarize a signed attestation | | 0g foundry … | optional plugin — hidden unless installed or --foundry |

Global flags: --network aristotle|galileo|local (default galileo), --rpc <url>, --private-key <hex>, --json.

Multi-language (the CLI is the universal surface)

TypeScript:

import { Storage } from "@foundryprotocol/0gkit-storage";
const s = new Storage({
  network: "galileo",
  privateKey: process.env.ZEROG_PRIVATE_KEY,
});
const { root } = await s.upload(new TextEncoder().encode("hi"));

Shell / any language (parse --json):

ROOT=$(0g storage put ./model.bin --network galileo --json | jq -r .root)
0g storage exists "$ROOT" --json

curl (compute is OpenAI-compatible — see @foundryprotocol/0gkit-compute):

0g infer -m "hello" --provider 0xPROVIDER --json | jq -r .output

Estimating & dry-run

0g estimate storage ./big-file.bin
0g estimate compute --prompt "What is 2+2?" --max-output 64
0g estimate da --bytes 4096
0g estimate contracts --abi ./MyContract.abi.json --address 0x... --method transfer --args '["0x...","1000"]'

Every write command also accepts --dry-run:

0g storage put ./file --dry-run
0g da publish ./blob --dry-run
0g infer -m "ping" --dry-run

--dry-run runs all estimation work (gas + fee + simulation) without broadcasting a single tx. The output is a structured DryRunResult envelope under --json, or a human [dry-run]-prefixed block by default.

Environment variables

ZEROG_NETWORK, ZEROG_RPC_URL, ZEROG_PRIVATE_KEY, ZEROG_BROKER_KEY, ZEROG_PROVIDER. Flags always override env; env overrides the preset default.

Escape hatch

The CLI is a thin, faithful wrapper. Every primitive package exposes .raw (or the underlying client) — drop down to @foundryprotocol/0gkit-storage, @foundryprotocol/0gkit-compute, @foundryprotocol/0gkit-da, @foundryprotocol/0gkit-attestation, @foundryprotocol/0gkit-chain, @foundryprotocol/0gkit-core directly whenever you outgrow a command. The toolkit is a help, never a cage.

License

MIT.