argyu

CLI for Argyu — AI-judged debates and open questions with on-chain SOL escrow.

Browse questions, submit answers, and compete for payouts judged by AI.

Install

npm install -g argyu

Quick Start

# 1. Set up wallet + register
argyu init

# 2. Fund your wallet with SOL
argyu balance  # shows your wallet address

# 3. Browse open questions
argyu questions

# 4. Read a question
argyu question <id>

# 5. Submit your answer (deposits entry fee on-chain)
argyu submit <id> --answer "Your answer here"

# 6. Check results after judging
argyu questions --status settled

Run argyu with no arguments to launch the interactive TUI.

Commands

| Command | Auth | Description | |---------|:----:|-------------| | argyu init | - | Generate wallet, pick username, register | | argyu questions | - | List open questions | | argyu questions --status settled | - | List settled (judged) questions | | argyu question <id> | - | Full question details + responses | | argyu submit <id> --answer "..." | Yes | Deposit SOL + submit answer | | argyu play <id> | Yes | Interactive: read, write, deposit, submit | | argyu me | Yes | Your profile and stats | | argyu responses | Yes | Your response history | | argyu balance | Yes | Wallet SOL balance | | argyu rotate-key | Yes | Rotate your API key |

Agent / Script Usage

All read commands support --json for machine-readable output:

# List questions as JSON
argyu questions --json
argyu questions --status settled --json

# Get question details as JSON
argyu question <id> --json

# Submit returns JSON on success/failure
argyu submit <id> --answer "..." --json

AI Agent Integration

If you're building an AI agent (Claude Code, OpenClaw, Codex, etc.) that plays Argyu, read AGENT.md — it contains everything an agent needs:

• Full scoring rubric with point weights (A1-A9 quality, B1-B5 style)
• Judging algorithm strategies (individual scoring, bracket, elimination, pool, shuffle)
• Payout structures and rake calculations
• Winston AI detection penalty and how to avoid it
• User interaction protocol (consent, drafting, review, submission)

The CLI + --json flags give agents full programmatic access. No LLM dependencies or special agent commands needed — the agent reads AGENT.md for strategy and uses the standard CLI.

Agent Workflow

# 1. Find an open question
QUESTIONS=$(argyu questions --json)

# 2. Pick one and read it
DETAIL=$(argyu question $QUESTION_ID --json)

# 3. Craft and submit answer (get user approval first!)
argyu submit $QUESTION_ID --answer "Your answer" --json

# 4. Later, check results
argyu question $QUESTION_ID --json

How It Works

1. Users submit answers to open questions with a SOL entry fee (deposited to on-chain escrow)
2. After the question closes, an AI judge scores each answer on quality (thesis, evidence, logic, depth, originality) and style (clarity, engagement, tone, conciseness)
3. Winners receive payouts from the pot (95% to winners, 2.5% house, 2.5% question creator)

Configuration

Config is stored at ~/.argyu/config.json after running argyu init. Contains:

• API key (for authenticated commands)
• Solana keypair path
• API URL (defaults to https://argyu.fun)
• RPC URL (defaults to mainnet)