robonet-workbench

使用 Robonet 的 MCP 服务器来构建、回测、优化和部署交易策略。提供 24 种用于加密货币和预测市场交易的专用工具：(1) 用于浏览策略、交易品种、指标、Allora 主题和回测结果的数据工具，(2) 用于生成策略想法和代码、优化参数以及通过 ML 预测进行增强的 AI 工具，(3) 用于测试历史数据策略性能的回测工具，(4) 用于 Polymarket 交易策略的预测市场工具，(5) 用于在 Hyperliquid 上进行实时交易的部署工具，(6) 用于信用管理的账户工具。使用场合：构建交易策略、回测策略、部署交易机器人、与 Hyperliquid 或 Polymarket 合作，或使用 Allora Network ML 预测增强策略。

## 概述（中文）

使用 Robonet 的 MCP 服务器来构建、回测、优化和部署交易策略。提供 24 种用于加密货币和预测市场交易的专用工具：(1) 用于浏览策略、交易品种、指标、Allora 主题和回测结果的数据工具，(2) 用于生成策略想法和代码、优化参数以及通过 ML 预测进行增强的 AI 工具，(3) 用于测试历史数据策略性能的回测工具，(4) 用于 Polymarket 交易策略的预测市场工具，(5) 用于在 Hyperliquid 上进行实时交易的部署工具，(6) 用于信用管理的账户工具。使用场合：构建交易策略、回测策略、部署交易机器人、与 Hyperliquid 或 Polymarket 合作，或使用 Allora Network ML 预测增强策略。

## 原文

# Robonet MCP Integration

## Overview

Robonet provides an MCP server that enables AI assistants to build, test, and deploy trading strategies. The server offers 24 tools organized into 6 categories: Data Access (8), AI-Powered Strategy Generation (6), Backtesting (2), Prediction Markets (3), Deployment (4), and Account Management (2).

## Quick Start

Load the required MCP tools before using them:

```
Use MCPSearch to select: mcp__workbench__get_all_symbols
Use MCPSearch to select: mcp__workbench__create_strategy
Use MCPSearch to select: mcp__workbench__run_backtest
```

After loading, call the tools directly to interact with Robonet.

## Tool Categories

### 1. Data Access Tools (Fast, <1s execution)

Browse available resources before building strategies:

- **`get_all_strategies`** - List your trading strategies with optional backtest results
- **`get_strategy_code`** - View Python source code of a strategy
- **`get_strategy_versions`** - Track strategy evolution across versions
- **`get_all_symbols`** - List tradeable pairs on Hyperliquid (BTC-USDT, ETH-USDT, etc.)
- **`get_all_technical_indicators`** - Browse 170+ indicators (RSI, MACD, Bollinger Bands, etc.)
- **`get_allora_topics`** - List Allora Network ML prediction topics
- **`get_data_availability`** - Check data ranges before backtesting
- **`get_latest_backtest_results`** - View recent backtest performance

**Pricing**: Most $0.001, some free. Use these liberally to explore.

**When to use**: Start every workflow by checking available symbols, indicators, or existing strategies before generating new code.

### 2. AI-Powered Strategy Tools (20-60s execution)

Generate and improve trading strategies:

- **`generate_ideas`** - Get AI-generated strategy concepts based on market data
- **`create_strategy`** - Generate complete Python strategy from description
- **`optimize_strategy`** - Tune parameters for better performance
- **`enhance_with_allora`** - Add Allora Network ML predictions to strategy
- **`refine_strategy`** - Make targeted code improvements
- **`create_prediction_market_strategy`** - Generate Polymarket YES/NO trading logic

**Pricing**: Real LLM cost + margin ($0.50-$4.50 typical). These are the most expensive tools.

**When to use**: After understanding available resources, use these to build or improve strategies. Always backtest after generation.

### 3. Backtesting Tools (20-40s execution)

Test strategy performance on historical data:

- **`run_backtest`** - Test crypto trading strategies
- **`run_prediction_market_backtest`** - Test Polymarket strategies

**Pricing**: $0.001 per backtest

**Returns**: Performance metrics (Sharpe ratio, max drawdown, win rate, total return, profit factor), trade statistics, equity curve data

**When to use**: After creating or modifying a strategy, always backtest before deploying. Use multiple time periods to validate robustness.

### 4. Prediction Market Tools

Build Polymarket trading strategies:

- **`get_all_prediction_events`** - Browse available prediction markets
- **`get_prediction_market_data`** - Analyze YES/NO token price history
- **`create_prediction_market_strategy`** - Generate Polymarket strategy code

**Pricing**: $0.001 for data tools, Real LLM cost + margin for creation

**When to use**: For prediction market trading strategies on Polymarket (politics, crypto price predictions, economics events)

### 5. Deployment Tools

Deploy strategies to live trading on Hyperliquid:

- **`deployment_create`** - Launch live trading agent (EOA or Hyperliquid Vault)
- **`deployment_list`** - Monitor active deployments
- **`deployment_start`** - Resume stopped deployment
- **`deployment_stop`** - Halt live trading

**Pricing**: $0.50 to create, free for list/start/stop

**Constraints**:
- EOA (wallet): Max 1 active deployment per wallet
- Hyperliquid Vault: Requires 200+ USDC in wallet, unlimited deployments

**When to use**: After thorough backtesting shows positive results. Never deploy without backtesting first.

### 6. Account Tools

Manage credits and view account info:

- **`get_credit_balance`** - Check available USDC credits
- **`get_credit_transactions`** - View transaction history

**Pricing**: Free

**When to use**: Check balance before expensive operations. Monitor spending via transaction history.

## Common Workflows

### Workflow 1: Create and Test New Strategy

```
1. get_all_symbols → See available trading pairs
2. get_all_technical_indicators → Browse indicators
3. create_strategy → Generate Python code from description
4. run_backtest → Test on 6+ months of data
5. If promising: optimize_strategy → Tune parameters
6. If excellent: enhance_with_allora → Add ML signals
7. run_backtest → Validate improvements
8. If ready: deployment_create → Deploy to live trading
```

**Cost**: ~$1-5 depending on optimization and enhancement

### Workflow 2: Enhance Existing Strategy

```
1. get_all_strategies (include_latest_backtest=true) → Find strategy
2. get_strategy_code → Review implementation
3. refine_strategy (mode="new") → Make targeted improvements
4. run_backtest → Test changes
5. If better: enhance_with_allora → Add ML predictions
6. run_backtest → Final validation
```

**Cost**: ~$0.50-2.00

### Workflow 3: Prediction Market Trading

```
1. get_all_prediction_events → Browse markets
2. get_prediction_market_data → Analyze price history
3. create_prediction_market_strategy → Build YES/NO logic
4. run_prediction_market_backtest → Test performance
5. If profitable: deployment_create → Deploy (when supported)
```

**Cost**: ~$0.50-5.00

### Workflow 4: Explore Ideas Before Building

```
1. get_all_symbols → Check available pairs
2. get_allora_topics → See ML prediction coverage
3. generate_ideas (strategy_count=3) → Get AI concepts
4. Pick favorite idea
5. create_strategy → Implement chosen concept
6. run_backtest → Validate
```

**Cost**: ~$0.50-4.50 (use generate_ideas to explore cheaply)

## Strategy Development Best Practices

### Start with Data Exploration

Always check availability before building:
- Use `get_data_availability` to verify symbol has sufficient history
- Check `get_allora_topics` if planning ML enhancement
- Review `get_all_technical_indicators` to know what's available

### Always Backtest

Never deploy without backtesting:
- Test on 6+ months of data minimum
- Use multiple time periods (train vs validation)
- Check metrics: Sharpe >1.0, max drawdown <20%, win rate 45-65%
- Compare performance across different market conditions

### Cost Management

Tools are priced in tiers:
1. **Data tools** ($0.001 or free) - Use liberally
2. **Backtesting** ($0.001) - Use frequently
3. **AI generation** (LLM cost + margin) - Most expensive
4. **Deployment** ($0.50) - One-time per deployment

**Cost-saving tips**:
- Use `generate_ideas` ($0.05-0.50) before `create_strategy` ($1-4)
- Check `get_latest_backtest_results` (free) before running new backtest
- Use `refine_strategy` ($0.50-1.50) instead of regenerating with `create_strategy`
- Review `get_strategy_code` (free) before modifying

### Strategy Naming Convention

Follow this pattern: `{Name}_{RiskLevel}[_suffix]`

Examples:
- `RSIMeanReversion_M` - Base strategy, medium risk
- `MomentumBreakout_H_optimized` - After optimization, high risk
- `TrendFollower_L_allora` - With Allora ML, low risk

Risk levels: H (high), M (medium), L (low)

## Technical Details

### Strategy Framework

Strategies use the Jesse trading framework with these required methods:
- `should_long()` - Check if conditions met for long entry
- `should_short()` - Check if conditions met for short entry
- `go_long()` - Execute long entry with posi