AI Chat Assistant for Crypto Trading Bots — VolatiCloud

Comparing four backtest results across three strategies usually means six tabs, a spreadsheet, and a fresh cup of coffee. Asking the VolatiCloud AI Chat Assistant "rank my last 10 backtests by Sharpe and tell me which strategy has the best risk-adjusted return" returns the answer in under a minute — drawn from your real account data, not generic trading advice. The assistant lives in a slide-out panel on every dashboard page, authenticates with your existing session, and can both answer questions and take actions.

What Is the AI Chat Assistant?

The AI Chat Assistant is an agentic AI panel embedded directly in the VolatiCloud dashboard. It opens in a slide-out panel from any screen, and it has live access to your organization's resources — strategies, bots, backtests, exchanges, performance data — through the same GraphQL API that powers the rest of the dashboard.

Unlike a generic chatbot, this assistant operates with org-scoped authorization. Every conversation is authenticated using your existing dashboard session. The assistant sees exactly what you see: scoped to your organization, respecting your role and permissions. It cannot access data from other organizations, and it cannot take actions your role doesn't permit.

The VolatiCloud MCP server works differently: it lets external AI tools — Claude Code, Claude Desktop, Cursor — connect to VolatiCloud via the Model Context Protocol bridge. The AI Chat Assistant is built into the product. No configuration, no separate API client, no external tool to install.

Agentic: It Can Act, Not Just Answer

The "agentic" part is the key distinction. A traditional chatbot replies with text and stops. An agentic assistant calls tools — queries, mutations, lookups — to accomplish tasks on your behalf.

Ask "What's my best-performing strategy this month?" and the assistant queries your backtest results, ranks them by Sharpe ratio and profit factor across your strategies, and returns a ranked answer backed by real data from your account.

Ask "Create a backtest for the RSI mean reversion strategy over the last 90 days on BTC/USDT" and the assistant initiates the run. It doesn't just tell you how to do it.

Every tool call the assistant makes is visible in the chat UI. A collapsible card surfaces the action (query name, inputs, output status) before the assistant's natural-language response. This transparency means you always know what the assistant did, not just what it said — useful both for trust and for debugging when an answer looks off.

What the Assistant Can Do

Because it shares the GraphQL API with the dashboard, the assistant can:

• Query your data — list strategies, bots, exchanges, backtest results, trade history
• Explain results — interpret a backtest's profit factor, Sharpe ratio, max drawdown, or win rate in plain English
• Compare options — rank strategies by performance metrics, compare runner configurations, evaluate exchange connections
• Trigger operations — initiate backtests, check bot status, start or stop bots (with your explicit confirmation on writes)
• Surface context — when you're viewing a specific strategy or bot, ask questions about it without specifying which one

The dashboard stops being a series of static screens and becomes a conversational interface to your trading data — particularly useful for the "I have a question that would take 4 clicks" cases that pile up over a trading week.

Org-Scoped Auth: Your Data Stays Yours

Security and data isolation are not an afterthought. The auth model is identical to the one used throughout VolatiCloud: every request the assistant makes goes through the same JWT verification and Keycloak-enforced permission checks that protect your organization's resources.

Concretely, this means:

• Your session, your scope: The assistant authenticates using your existing dashboard session. It inherits your role (admin, member, viewer) and cannot escalate privileges.
• Organization isolation: If you're in Org A, the assistant cannot access Org B's data even if you're a member of both. Conversations are scoped to the currently active organization.
• Viewer-safe: A viewer-role user can ask the assistant to retrieve data, but it will refuse to execute write operations (creating bots, starting backtests) because your role doesn't permit those actions.
• Conversation history is private: Past conversations are stored under your organization's namespace. Other members of your organization cannot read your conversations.

The org-scoped model also makes the assistant context-aware. When you navigate to a specific strategy page and open the chat panel, the assistant knows which strategy you're viewing. You can ask "What are the RSI parameters in this strategy?" without specifying which one — it understands the current context.

Persistent Conversation History

Each chat session creates a named conversation. Conversations persist across sessions. Close the panel, come back tomorrow, and your conversation history is intact.

The conversation list view shows all your past sessions, ordered by recency. Each conversation shows a title (auto-generated from the first message) and timestamp. You can switch between conversations, continue where you left off, or start a new chat when switching context.

This is useful for iterative work. A typical workflow might look like:

1. Monday: You ask the assistant to analyze your last three backtest runs and identify the parameter that most affects drawdown
2. Tuesday: You reopen the same conversation, apply the suggested parameter change in the Strategy Builder, run a new backtest
3. Wednesday: You come back to compare before/after results in the same conversation thread, with context intact

Rather than starting from scratch each session, you build up a working thread with full history.

Practical Use Cases

Strategy Research

When you're building a new strategy in the Visual Strategy Builder, the assistant can help you think through indicator combinations and their historical behavior. Ask it to explain what a crossover between a 12-period EMA and 26-period EMA signals, or whether adding an RSI filter to your existing trend-following logic makes conceptual sense.

More practically, once you've run a backtest, you can ask the assistant to interpret the results:

"My backtest shows a profit factor of 1.4 with a Sharpe of 0.8 and a max drawdown of 23%. Is this worth deploying?"

The assistant won't just recite definitions. It'll put the numbers in context — a 23% max drawdown is substantial for a strategy targeting 15% annual returns, and it'll tell you that.

Operational Questions

Not everything requires deep analysis. Sometimes you just need quick answers:

• "How many bots do I have running right now?"
• "Which exchange is connected to my BTC/USDT bot?"
• "What was the last trade my ETH scalper made?"
• "Is my Binance API key still valid?"

These are the kinds of questions that normally require clicking through multiple screens. The assistant handles them in one message.

Backtest Comparison

After running multiple backtests with different parameter sets, comparing results by hand is tedious. The assistant can pull all backtest results for a strategy, extract the key metrics (profit factor, Sharpe, win rate, max drawdown, trade count), and present them as a ranked comparison. You identify which parameter combination performs best without opening each result individually.

This pairs naturally with hyperparameter optimization: run an optimization pass, then use the assistant to interpret which parameter set won and why.

Pre-Deployment Checklist

Before moving a bot from paper trading to live, use the assistant as a sanity check:

"Review my RSI mean reversion bot setup — is there anything I should verify before switching to live trading?"

The assistant checks the configuration, looks at recent backtest performance, verifies the exchange connection is active, and surfaces obvious issues: a stoploss percentage tighter than ATR-justified, or a timeframe mismatch between the backtest config and the live bot config. This pairs well with the paper-trading-to-live framework — let the assistant pre-screen, then run through the framework yourself.

AI Chat Assistant vs MCP Server

VolatiCloud has two AI-powered surfaces, and they serve different workflows:

	AI Chat Assistant	MCP Server
Location	Inside the dashboard	Your editor (Claude Code, Cursor, Claude Desktop)
Auth	Your dashboard session	OAuth PKCE through Keycloak
Setup	Zero — already there	Paste a JSON config block, sign in once
Best for	Operational questions, result analysis	Strategy code editing, IDE-driven workflows
Conversation history	Persisted in VolatiCloud per organization	In your editor's session
Actions	Full GraphQL access (read + write)	Full GraphQL access (read + write)

The two interfaces are complementary. Use the in-dashboard assistant when you're analyzing data or making operational decisions inside the browser. Use the MCP server when you want an AI coding assistant in your editor to help write and iterate on strategy logic, or when you prefer working from a terminal. The same authorization rules apply on both surfaces.

For the MCP setup walkthrough, see Connect Claude to your trading bots. For practical AI workflows, see AI crypto trading workflows with Claude.

Getting Started

The AI Chat Assistant is available to all VolatiCloud users. Open the dashboard at console.volaticloud.com and look for the chat icon in the navigation bar. The panel slides in from the right side of any screen.

A few suggestions for getting the most out of it:

Start with questions about your data. The assistant is strongest when it can pull real data from your account and reason about it. Generic questions about trading theory will work, but specificity gets much better results.

Name your conversations. When you start a new thread for a specific project (researching a new strategy, preparing to deploy a bot), give it context in your first message. The assistant will use that context throughout the conversation.

Check the tool call cards. Every action the assistant takes is shown as a collapsible card in the chat. Reviewing these helps you understand what data the assistant used to form its answer, and it's a good habit for validating responses.

Use it iteratively. The persistent history makes iterative workflows natural. Start a thread when you're researching a strategy, and keep adding to it as you backtest, optimize, and eventually deploy.

---

The AI Chat Assistant is live now. Log in to console.volaticloud.com and open the chat panel to try it. If you have feedback or feature requests, the assistant itself is a reasonable place to report what's not working — it surfaces context-aware issues better than a generic feedback form.

For broader context, see real-time bot monitoring and strategy versioning and forking for the underlying features the assistant works with.