ClawPoly is a local-first Polymarket trading bot focused on practical execution for small-to-mid bankrolls.
It combines event mispricing (value_betting) + MM-lite filtering, then gates every trade with risk controls and Kelly sizing.

• 2026 High-Frequency Async Stack:
  • ⚡ Full Asynchronous Execution: Parallel market scanning and orderbook analysis using asyncio for maximum speed.
  • 🏎️ fast_loop: HF Momentum / News Scalping — 100ms scans + front-running Binance/Oracle feeds. (ROI: 8–15%/mo).
  • 🤖 llm_catalyst_sniper: AI-Driven Catalyst Sniping — Using LLMs to front-run news and sentiment shifts.
  • 💧 market_making: Automated Market Making (AMM) — Capturing spreads + liquidity rewards. (ROI: 1–3%/mo).
  • 🕸️ intra_market_arb: Intra-market Arbitrage — Exploiting mathematical violations within a single event.
  • 🌦️ weather_trader: Weather Probability Edge — Trading temperature ranges using real-time forecast data.
• Whitelist Execution Guardrails:
  • Optional fee/rebate-only markets (feesEnabled)
  • Minimum spread threshold before execution
• Risk Engine:
  • Conservative/Moderate/Aggressive profiles
  • Fractional Kelly sizing + max concurrent limits + stop-loss
• Execution Safety Layer:
  • paper -> live mode switch
  • kill switch (immediate live stop + cancel-all in loop)
  • max loss/day guard + slippage cap + cooldown
• Terminal Experience:
  • Rich CLI output
  • Optional live TUI mode for scan/trade cycles
• 24/7 Autonomous Trading:
  • Continuous trading loops with real-time monitoring
  • OpenClaw Gateway integration for AI-powered decision making
  • WebSocket support for real-time AI scoring
• Polymarket API Integration:
  • Gamma API (events/markets)
  • CLOB API (orderbook/prices/trading)
  • Data API (positions/P&L)

flowchart TD
    A[Fetch Active Events/Markets] --> B[Strategy Scan]
    B --> C[value_betting + MM-lite signals]
    C --> D[Whitelist Filters]
    D -->|fee/rebate + min spread| E[Risk Manager]
    E -->|kelly + limits + stop-loss| F[Execute or Dry-Run]
    F --> G[Trade Log + TUI Update]

ClawPoly supports continuous, autonomous trading operations with real-time monitoring and AI-powered decision making through OpenClaw integration.

# Start 24/7 trading with live TUI monitoring
python main.py run --risk conservative --live --tui

# Background operation (no TUI)
python main.py run --risk moderate --live

ClawPoly integrates with OpenClaw AI assistant for enhanced trading capabilities:

• Real-time AI Scoring: WebSocket connection to OpenClaw Gateway for live AI analysis
• Voice Commands: Control trading via voice through OpenClaw's voice wake/talk modes
• Multi-channel Monitoring: Receive alerts and updates via WhatsApp, Telegram, Discord
• Skill-based Commands: Use natural language to query positions, scan markets, and manage trades

• Kill Switch: Emergency stop with python main.py kill-switch --on
• Real-time Stats: python main.py stats for P&L tracking
• Position Monitoring: python main.py balance for current holdings
• Risk Limits: Automatic daily loss caps and position limits
• Health Checks: Built-in monitoring for API connectivity and wallet balance

• Python 3.9+
• Polymarket wallet + USDC on Polygon

1. Clone and install dependencies

git clone https://github.com/OpenKrab/ClawPoly.git
cd ClawPoly
pip install -r requirements.txt

1. Configure environment

cp .env.example .env

1. Run tests

python -m pytest tests/ -q -p no:asyncio

# Show config
python main.py config

# Browse active markets
python main.py markets -n 20

# One-shot scan (no trades)
python main.py scan -n 20

# Continuous dry-run with live TUI
python main.py run --risk moderate --tui

# Live trading (real money)
python main.py run --risk conservative --live --tui

python main.py --help
python main.py balance
python main.py config
python main.py markets --search "bitcoin"
python main.py scan --strategy value_betting -n 30
python main.py run --strategy weather_trader --strategy fast_loop --mode live --tui
python main.py history --last 50
python main.py stats
python main.py provider-test
python main.py kill-switch --status
python main.py kill-switch --on --reason "manual stop"
python main.py cancel-all
python main.py run --risk moderate --max-trades 3 --tui

• ENABLED_STRATEGIES=value_betting,market_making
• ONLY_FEE_REBATE_MARKETS=true
• MIN_SPREAD=0.02
• MIN_SIGNAL_CONFIDENCE=0.25
• FEE_AWARE=true
• STRATEGY_SCAN_CANDIDATE_LIMIT=600

Public strategy aliases now map to the legacy internal ids automatically. Examples: For 4-layer model routing, also configure:

• AI_PROVIDER=claude (supported: claude, codex, openclaw, opencode, kilocode)
• AI_PROVIDER_MODEL= (optional label used in signal metadata)
• AI_PROVIDER_LIVE=false (set true to call real provider APIs)

These defaults are tuned for the real-world pattern of:

• event mispricing selection
• MM-lite opportunities in liquid, fee/rebate-eligible markets
• avoiding low-quality, low-spread fills

Example .env baseline:

PRIVATE_KEY=your_private_key_here
FUNDER_ADDRESS=your_wallet_address_here
SIGNATURE_TYPE=0

RISK_LEVEL=conservative
ENABLED_STRATEGIES=value_betting,market_making
AI_PROVIDER=claude
AI_PROVIDER_MODEL=
AI_PROVIDER_LIVE=false
AI_PROVIDER_TIMEOUT_SEC=8
AI_PROVIDER_MAX_LIVE_CALLS=25
STRATEGY_PROVIDER_OVERRIDES=weather_trader:claude,fast_loop:kilocode
STRATEGY_MODEL_OVERRIDES=
STRATEGY_SCAN_CANDIDATE_LIMIT=600

TRADING_MODE=paper
MAX_DAILY_LOSS_USD=25
DAILY_LOSS_PROXY_PCT=1.0
MAX_SLIPPAGE_PCT=0.02
TRADE_COOLDOWN_SEC=20
MAX_POSITION_USD=0
KILL_SWITCH_STATE_PATH=execution_state.json

ANTHROPIC_API_KEY=
AI_CLAUDE_BASE_URL=https://api.anthropic.com/v1/messages

OPENAI_API_KEY=
AI_CODEX_BASE_URL=https://api.openai.com/v1

OPENCLAW_API_KEY=
AI_OPENCLAW_BASE_URL=ws://127.0.0.1:18789
AI_PROVIDER_MODEL=

OPENCODE_API_KEY=
AI_OPENCODE_BASE_URL=https://opencode.ai/zen/v1

KILOCODE_API_KEY=
AI_KILOCODE_BASE_URL=https://openrouter.ai/api/v1

SCAN_INTERVAL=60
MIN_LIQUIDITY=500
MIN_VOLUME=1000

ONLY_FEE_REBATE_MARKETS=true
MIN_SPREAD=0.02
MIN_SIGNAL_CONFIDENCE=0.25
FEE_AWARE=true

See .env.example for the full set of options.

Live provider routing:

• claude: uses Anthropic Messages API (ANTHROPIC_API_KEY)
• codex: uses OpenAI-compatible API (OPENAI_API_KEY + AI_CODEX_BASE_URL)
• openclaw: OpenClaw Gateway integration with WebSocket support for real-time AI scoring (OPENCLAW_API_KEY + AI_OPENCLAW_BASE_URL=ws://127.0.0.1:18789 + explicit AI_PROVIDER_MODEL or strategy model override). Supports both REST API (http/https URLs) and WebSocket connections (ws/wss URLs) to OpenClaw Gateway.
• opencode: uses OpenCode Zen API (OPENCODE_API_KEY + AI_OPENCODE_BASE_URL=https://opencode.ai/zen/v1)
• kilocode: uses OpenAI-compatible API (KILOCODE_API_KEY + AI_KILOCODE_BASE_URL, fallback to OPENAI_API_KEY)
• On API errors/timeouts or missing keys, strategy falls back to local deterministic scoring
• Use python main.py provider-test to verify whether mode is truly live

• main.py - CLI entrypoint
• config/settings.py - env loading, risk profiles, runtime settings
• src/scanner.py - scan loop, whitelist checks, execution orchestration
• src/client.py - Polymarket CLOB wrapper
• src/data/ - markets/orderbook/positions/history APIs
• src/data/db.py - SQLite schema + stats queries
• src/strategies/value_betting.py - event mispricing logic
• src/strategies/market_making.py - MM-lite spread/liquidity logic
• src/strategies/penny_sniper.py - close-to-expiry execution
• src/strategies/sports_arb.py - guaranteed math edge sports logic
• src/ai/provider.py - provider router (claude/kilocode/codex/openclaw with WebSocket support)
• src/risk/manager.py - risk checks and portfolio gates
• src/risk/kelly.py - Kelly sizing helpers
• src/utils/display.py - rich display tables
• src/utils/tui.py - live terminal monitoring view
• SKILL.md - OpenClaw skill integration specification
• tests/ - strategy + risk tests

• Live trading is off by default (dry-run first)
• Use kill switch for emergency stop: python main.py kill-switch --on
• Keep private keys only in .env (already gitignored)
• Use conservative profile until execution behavior is stable
• Respect liquidity/spread filters to reduce slippage risk

python -m pytest tests/ -q -p no:asyncio

PRs are welcome. Please include tests for behavior changes, especially in:

• strategy signal generation
• whitelist/risk gating
• execution safety paths

Built for the Lobster Way 🦞

ClawPoly integrates seamlessly with OpenClaw AI assistant for enhanced trading capabilities:

1. Install OpenClaw: npm install -g openclaw@latest
2. Configure skill in ~/.openclaw/openclaw.json:

{
  "skills": {
    "entries": {
      "clawpoly": {
        "enabled": true,
        "apiKey": { "source": "env", "provider": "default", "id": "CLAUPOLY_PRIVATE_KEY" },
        "env": {
          "CLAUPOLY_PRIVATE_KEY": "your_private_key_here"
        }
      }
    }
  }
}

Once integrated, you can control ClawPoly through OpenClaw using natural language:

• "Show me current positions on Polymarket"
• "Scan for arbitrage opportunities"
• "Start trading with moderate risk"
• "What's my P&L today?"
• "Stop all trading immediately"

For real-time AI analysis, configure WebSocket connection:

# In .env
AI_OPENCLAW_BASE_URL=ws://127.0.0.1:18789
AI_PROVIDER=openclaw

This enables live AI scoring through OpenClaw Gateway for enhanced trading decisions.