LP Evil Panda

Meteora DLMM liquidity provision strategy implementing Evil Panda, powered by the LP Agent API. This repo contains the bot core and is intended to be used as an OpenClaw-managed plugin/integration for running the strategy.

What it does

1. Scans for pumping meme coins on Meteora with good volume, safety checks (Jupiter token safety, freeze/mint authority), and organic activity filters
2. Enters SOL-sided SPOT DLMM positions with a wide range (-80% to -90%) to earn fees from volatility
3. Monitors positions on 15-minute candles using technical indicators (Bollinger Bands, RSI(2), MACD)
4. Exits when 2+ confluence signals fire, or after 24h hard stop-loss if out of range with negative P&L
5. Alerts via Telegram on entry, exit, errors, and startup

Architecture

Current code path:

Scanner ──→ Executor (zap-in) ──→ Position Manager ──→ Executor (zap-out)
   │              │                       │                    │
   ▼              ▼                       ▼                    ▼
LP Agent API   Signer (OWS/local)   GeckoTerminal       LP Agent API
+ Jupiter      + Jito bundles        15m OHLCV           + Jito bundles

Intended operating model:

OpenClaw ──→ scheduled scan / heartbeat monitor ──→ bot core in this repo

Priority-based polling in the current implementation:

• Exit monitoring: every 1 minute (high priority)
• Pool scanning: every 5 minutes (lower priority)

Quick start

# Clone and install
git clone https://github.com/lpagent/lp-evil-panda.git
cd lp-evil-panda
yarn install

# Configure
cp .env.example .env
# Edit .env — at minimum set LP_AGENT_API_KEY and a wallet option

# Run locally (standalone loop)
yarn dev

# Run a single scan tick
yarn scan:once

# Run a single monitor/heartbeat tick
yarn monitor:once

For production use, install this repo as an OpenClaw plugin and trigger lp-scan on cron + lp-monitor on heartbeat cadence.

Get your LP Agent API key

1. Go to portal.lpagent.io
2. Create an account and generate an API key
3. Set LP_AGENT_API_KEY in your .env file

Wallet options

Setting up OpenWallet (OWS)

OpenWallet is the recommended signing method — your bot never touches raw private keys. Transactions are signed through a policy-gated CLI, so you can restrict what your agent is allowed to do.

Install OWS

Use the official links:

• OpenWallet
• OpenWallet docs
• OpenClaw (orchestrator / assistant)

OpenClaw is the intended orchestration layer for this strategy. OpenWallet / ows is still the signing layer used by the bot.

If your OpenClaw workflow installs OWS for you, verify it with:

ows --version

Create a wallet for your bot

# Create a new Solana wallet
ows create wallet --name agent-treasury --chain solana

# View the wallet address (fund this with SOL)
ows address --wallet agent-treasury --chain solana

# Set in your .env
echo 'OWS_WALLET_NAME=agent-treasury' >> .env

The bot automatically detects OWS on your PATH and uses it for all transaction signing. If OWS is unavailable, it falls back to local keypair signing.

When running as OpenClaw plugin, lp-start / lp-setup can attempt to install OWS automatically via npm i -g @open-wallet-standard/core when OWS_WALLET_NAME is configured but ows is missing.

OpenClaw orchestration model

This repo should be treated as the strategy core that OpenClaw drives.

OpenClaw-facing surfaces added in this repo:

• plugin entry: src/openclaw-plugin.ts
• plugin manifest: openclaw.plugin.json
• one-shot scan runner: src/scan-once.ts
• one-shot monitor runner: src/monitor-once.ts

Registered OpenClaw commands:

• lp-start — bootstrap setup check (env + signer + balance) and attempt OpenWallet auto-install if OWS_WALLET_NAME is set but ows is missing
• lp-setup — alias of lp-start
• lp-scan — run one entry scan / execution cycle
• lp-monitor — run one monitor / exit cycle
• lp-status — show tracked local state

Recommended OpenClaw responsibilities:

• trigger lp-scan on a cron schedule
• trigger lp-monitor on a heartbeat cadence
• surface alerts / progress updates
• restart or re-run monitoring flows after failures

Suggested cadence:

• heartbeat / monitor: every 60s
• cron / scan: every 5m

These match the strategy defaults:

• POLL_INTERVAL_MS=60000
• SCAN_INTERVAL_MS=300000

To avoid overlapping jobs, discrete runners use a repo-local lock file: .openclaw-orchestrator.lock.

Configuration

All parameters are configurable via .env. See .env.example for the full list.

Strategy parameters

Scanner filters

Exit signals (15m timeframe)

The bot exits when 2 or more of these signals fire simultaneously:

• Bollinger Bands: Close above upper band (period=20, stdDev=2)
• RSI(2): Value > 90 (extremely overbought on 2-period RSI)
• MACD: First green histogram bar after red (12/26/9)

LP Agent API endpoints used

This bot demonstrates the full LP Agent API workflow:

Project structure

src/
  index.ts              Standalone entrypoint
  orchestrator.ts       Standalone loop + one-shot orchestration runners
  app-context.ts        Shared bootstrap for config/clients/state
  cycles.ts             Reusable scan and monitor cycle logic
  openclaw-plugin.ts    OpenClaw plugin commands (lp-scan, lp-monitor, lp-status)
  scan-once.ts          One-shot scan entrypoint
  monitor-once.ts       One-shot monitor entrypoint
  run-lock.ts           Lock file protection for discrete jobs
  config.ts             Environment config loader
  lp-agent-client.ts    LP Agent API client (typed)
  gecko-client.ts       GeckoTerminal OHLCV client (rate-limited)
  jupiter-client.ts     Jupiter token safety API
  signer.ts             Signing abstraction (OWS + local keypair)
  scanner.ts            Pool discovery + safety filters
  analyzer.ts           Technical indicator computation (BB, RSI, MACD)
  position-manager.ts   Position evaluation and exit decisions
  executor.ts           Zap-in/out execution with retry logic
  state.ts              JSON file persistence
  telegram.ts           Telegram Bot API notifications
test/
  analyzer.test.ts      Exit signal detection tests
  executor.test.ts      Retry logic and failure handling tests
  position-manager.test.ts  Stop-loss and stale data tests
  signer.test.ts        Wallet signing tests
  state.test.ts         State persistence tests

Testing

yarn build         # Compile TypeScript
yarn test          # Run all tests
yarn test:watch    # Watch mode

OpenClaw plugin packaging notes

This repository now exposes an OpenClaw plugin extension via package.json:

• openclaw.extensions: ./dist/openclaw-plugin.js
• manifest file: openclaw.plugin.json

After yarn build, OpenClaw can load the built extension entry.

Roadmap

See TODOS.md for planned features.

License

MIT