Agent

Reversion has two AI agents, both built into the web app and both connectable via MCP/API:

• Auto-Trading Agent — an autonomous trader that watches markets, validates setups with chart and news alerts and analysis, and executes trades with built-in risk management.
• Algorithm Agent (MCP) — an algorithm builder assistant 60+ tools for building algorithms, backtesting, and trading through natural language.

---

Auto-Trading Agent

The auto-trading agent lives inside the Reversion web app. You chat with it, configure your risk limits, and it handles the rest — watching markets, analyzing charts, placing trades with mandatory stop losses, and managing positions autonomously.

1. Open Agent Mode

   Click Agent in the header mode selector. You’ll see a three-column layout:

   • Left: Live state sidebar — account equity, active trades, watched markets, risk utilization
   • Center: Chat interface — talk to the agent in natural language
   • Right: Alert threads — autonomous actions the agent takes when markets hit your levels

2. Connect Your LLM API Key

   The agent needs an LLM provider to reason. Open the model settings and add your API key for one of the supported providers:

   • Anthropic (Claude)
   • OpenAI (GPT)
   • Google (Gemini)
   • xAI (Grok)
   • Groq

   Your API key is encrypted and stored securely — it’s never exposed to other services.

   Note

   Claude (Anthropic) is recommended for the best chart analysis and trading reasoning. The agent uses vision to read candlestick charts, so model quality matters.

3. Configure Risk Settings

   Open the risk config panel. These settings control how much the agent can risk:

   Setting	What It Does	Suggested Start
   Max risk per trade	Maximum USD you can lose on any single trade	$10-50
   Max portfolio risk	Total USD at risk across all open trades	$50-200
   Max drawdown %	Locks out new trades if equity drops this much from peak	10-20%
   Min R:R	Minimum reward-to-risk ratio for any trade	1.0

   Caution

   Start conservative. You can always increase limits once you’re comfortable with how the agent trades. The agent literally cannot place a trade that violates these limits — they’re enforced at the system level, not by prompting.

4. Choose Your Operating Mode

   Mode	How It Works
   Copilot	The agent proposes trades, and you approve or reject each one via a popup. Best for learning how the agent thinks.
   Full Autonomy	The agent executes trades immediately without asking. You can disconnect and it keeps running.

   Note

   Start with Copilot mode. Watch how the agent evaluates setups and what trades it proposes. Switch to Full Autonomy once you trust its judgment.

5. Start Trading

   Ask the agent to scan markets and set up watchers:

   "Scan BTC and ETH on the 4h timeframe, set up watchers at key levels"

   The agent will:

   1. Analyze charts across multiple timeframes using vision
   2. Identify key support/resistance levels and market structure
   3. Set price alerts at those levels with proposed trade setups
   4. When a level is hit — an alert thread spawns automatically
   5. The alert thread validates the setup against current conditions
   6. If valid, it executes the trade with your risk settings (mandatory SL, position size derived from your max risk)
   7. After entry, management alerts auto-trigger to tighten stops, take profit, or trail

   You’ll see alert threads appear in the right panel as the agent works autonomously.

6. Monitor Your Trades

   The live state sidebar shows everything in real-time:

   • Account: equity, margin, available balance
   • Active trades: entry, stop loss, target, unrealized P&L, risk per trade
   • Watched markets: price levels the agent is monitoring
   • Risk utilization: how much of your risk budget is in use
   • Drawdown: current drawdown from peak equity

   Every trade has a mandatory stop loss placed on the exchange. The agent can tighten stops and trail into profit, but can never widen them — risk only decreases after entry.

Note

You can inject messages into running alert threads if you want to override the agent’s decision. The agent also maintains trading notes — persistent memory it uses across all threads to track its market thesis and strategy reasoning.

---

Algorithm Agent (MCP)

The Algorithm Agent is built into the Reversion web app — click Agent in the header and chat with it directly. You can also connect it to external apps like Claude Desktop, Cursor, or VS Code via the Model Context Protocol. Either way, you get 60+ tools for building algorithms, backtesting, optimizing parameters, and trading through natural language.

Using in the Web App

Click Agent in the header to open the chat interface. The Algorithm Agent is ready to use immediately — no setup required. Just describe what you want:

"Build a mean reversion strategy for ETH using Bollinger Bands"
"Backtest my algorithm over the last 3 months and plot the equity curve"
"Optimize RSI parameters with 20 trials"

Connecting via MCP (Optional)

If you prefer working in Claude Desktop, Cursor, VS Code, or another MCP-compatible app, follow the steps below to connect the Algorithm Agent externally.

1. Create an API Token

   Sign in to reversion.trade with your wallet (MetaMask, Rabby, or Phantom). Navigate to the API Tokens section from the header.

   • Name: Give it a descriptive name (e.g., reversion-agent)
   • Scopes: Select read, trade, and algo for full access
   • Expiry: Optional — set an expiry date or leave it permanent

   Save the token (rvt_...) — it’s only shown once.

   Note

   Your private keys are never exposed. The platform creates a separate API wallet for executing trades, and your token authenticates against it. You can revoke tokens at any time.

2. Configure Your MCP Client

   Add the Reversion MCP server to your AI app. The simplest setup uses the npm package:

   {% tabs %} {% tab title=“Claude Desktop” %} Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

   {
     "mcpServers": {
       "reversion": {
         "command": "npx",
         "args": ["-y", "reversion-mcp"],
         "env": {
           "REVERSION_API_TOKEN": "rvt_your_token_here",
           "REVERSION_BACKEND_URL": "https://api.reversion.trade"
         }
       }
     }
   }

   {% endtab %}

   {% tab title=“Cursor” %} Edit .cursor/mcp.json in your project or home directory:

   {
     "mcpServers": {
       "reversion": {
         "command": "npx",
         "args": ["-y", "reversion-mcp"],
         "env": {
           "REVERSION_API_TOKEN": "rvt_your_token_here",
           "REVERSION_BACKEND_URL": "https://api.reversion.trade"
         }
       }
     }
   }

   {% endtab %}

   {% tab title=“VS Code” %} Edit .vscode/mcp.json:

   {
     "servers": {
       "reversion": {
         "command": "npx",
         "args": ["-y", "reversion-mcp"],
         "env": {
           "REVERSION_API_TOKEN": "rvt_your_token_here",
           "REVERSION_BACKEND_URL": "https://api.reversion.trade"
         }
       }
     }
   }

   {% endtab %} {% endtabs %}

   Restart your app after saving. The Reversion tools should appear in the tool list.

3. Verify It Works

   Try asking your agent:

   “Generate a 4h chart for BTC/USDC with RSI and Bollinger Bands”

   If everything is configured correctly, you’ll see a candlestick chart with indicators rendered inline.

Caution

If you give the agent a token with the trade scope, it can place real orders with real money. Start with read + algo scopes if you just want analysis and backtesting without live trading risk.

---

What’s Next?

Want to…	Go to
See full auto-trading agent reference	Agent Features → Auto-Trading Agent
See all 60+ Algorithm Agent tools	Agent Features → Algorithm Agent