2 AI components Individually activatable Real-time web search

Claude AI integration — optional AI check layer

FuturesBot offers an optional integration with Anthropic Claude AI as an additional check layer. 2 independent AI components: trade check before entry and exit check at the time exit. All optional, all configurable, not an investment recommendation.

Note: Claude AI is not a fixed component of FuturesBot, but an optional extension (off by default). You need your own Anthropic API key (paid, directly with Anthropic) and you have to top up credit. FuturesBot works fully without AI — all 14 strategies, risk management and protection systems work independently.
FuturesBot with Claude AI integration enabled — live validation visible

Claude AI directly in the bot: trade check and exit check — visible in the user interface.

🤖 Artificial intelligence

What the AI check does

If the AI check is enabled, Claude checks new trades before execution and can block them — not as a crystal ball, but as additional check layer according to your rules. Whether this improves your results is not proven — first check thoroughly in paper mode.

Web search before entry
The main basis of the trade check is the bot's technical market data. Additionally, Claude can search the web for acute extreme events look for — flash crashes, liquidation cascades, exchange outages — before the trade is executed.
No crystal ball — an additional check
Claude AI does not predict the future. It evaluates the bot's market data and can, via web search, acute extreme events include — and may also overlook or misjudge something.
Completely optional
The bot works entirely without AI. All 14 strategies, risk management and protection mechanisms work completely independently. Claude AI is an optional, paid extension — you set it up yourself with your own Anthropic account and pay only for what you use.
Your own API key
You create your own your own Anthropic account at console.anthropic.com, top up credit (min. $5) and receive your API key. No middlemen — you pay Anthropic directly, only for what you use. FuturesBot charges no additional costs for the AI function.
--- AI check active ---

🤖 Claude AI: Checking trade signal...
WebSearch: Searching for risks...

📰 WARNING: reports of
unusual liquidations
($xxxM liquidated in 1h)

📊 Funding rate: +0.15% (extreme)
📉 Momentum (1–5 min): choppy

❌ BLOCK (Confidence: xx/100)
Reason: extreme liquidations +
choppy momentum. Wait.

Example display with fictitious values — not a real signal, not a result

The 2 AI components in the Detail

Each component can be enabled individually. Use one or both — depending on your needs and budget.

1

Trade validation

As the last check before the order, Claude evaluates the signal based on the bot's technical market data (RSI, MACD, EMA, volume, ATR, ADX, funding) and the signal context. WebSearch is intended only for acute extreme events (flash crash, exchange outage, large liquidations).

Model: Claude Sonnet Pay-per-use costs WebSearch active

How it works: The response is structured: YES or NO + confidence (0–100) + reasoning. On NO, at a confidence below CLAUDE_MIN_CONFIDENCE (default 70); on timeout or error the trade is not executed.

Activate: ENABLE_CLAUDE_VALIDATION=true + ENABLE_CLAUDE_TRADING_EXECUTOR=true (off by default)

--- Claude AI trade validation ---
Model: Claude Sonnet
Context: Bot market data + web search

Signal: LONG ETH/USD
Score: xx/100

--- Analysis ---
📈 RSI(14): xx · MACD: bullish
📊 Funding neutral (0.01%)
📈 Momentum (1–5 min): clear
📊 Volume: x,x× average
🔍 WebSearch: no extreme events

✅ Result: APPROVAL
Confidence: xx/100
Reasoning: No blocking reasons according to
found by your rules —
not an investment recommendation.

Cost: $x.xxx (estimated) | Duration: x,x s

Example display with fictitious values — not a real signal, not a result
2

Exit check on time exit (optional)

If a position reaches its maximum holding time without a notable profit, the bot asks Claude for the short-term market direction. If the assessment points in the direction of the position with sufficient confidence, the time exit is postponed by a few minutes (at most 3×); otherwise the position is closed. Postponing can increase the loss. Not an investment recommendation.

Model: Claude Sonnet Pay-per-use costs WebSearch active

Process: Holding time reached, profit below 0,5 % and pre-check passed (3 of 5 checks) → Claude estimates via WebSearch (news and sentiment of the last 1–2 hours, BTC correlation) the direction of the next 15–30 minutes: LONG, SHORT or NEUTRAL. A shift only occurs if the confidence is above CLAUDE_EXIT_MIN_CONFIDENCE (default 70) in the position direction; in case of NEUTRAL, opposite direction, timeout or error, the position is closed. Not an entry filter.

Activate: ENABLE_CLAUDE_EXIT_VALIDATION=true (off by default; independent of the master switch)

--- Exit check (time exit) ---
Model: Claude Sonnet
Position: LONG SOL/USD · holding period reached
PnL: ±x,x %

--- WebSearch (last 2h) ---
📰 SOL Ecosystem: New DEX volume ATH
📊 BTC stable above $68k
💹 Altcoin sentiment: bullish
📈 SOL/BTC ratio rising

--- Assessment (15–30 min) ---
→ LONG
Confidence: xx/100
Reasoning: Bullish ecosystem
Momentum + rising BTC correlation.
→ Time exit postponed (1/3).

Cost: $x.xxx (estimated) | Duration: x,x s

Example display with fictitious values — not a real signal, not a result

Costs & Models

You pay by usage directly to Anthropic — tokens and web searches. FuturesBot charges nothing for this.

Haiku Sonnet Opus
Speed Fastest Fast Slower
Cost (relative) low medium high
Default setting — Trade + Exit —
Monthly costs:
The costs depend on the number of trades, the model and web search (Anthropic charges extra for web searches).
With many trades per day, several dozen euros per month can accrue — what counts is your billing in the Anthropic console.
Claude is only asked when a trade has passed all other checks, or, for the time exit, when the pre-check (3 of 5 checks) is met.
Token logging:
The bot estimates the token costs per call (fixed price, USD; web searches and deviating model prices not included) and writes them to:
logs/claude_tokens.csv
Your billing in the Anthropic console is authoritative. Input and output tokens as well as the estimated costs are recorded.

Setup in 6 steps

Setup step by step, without programming skills – then check in paper mode first.

1
Create an Anthropic account
Visit console.anthropic.com and create a free account.
2
Create API key
Under Settings → API Keys create a new key. Store it safely — it is shown only once.
3
Top up balance
Minimum $5 to start — how long that lasts depends on the number of trades, model and web search. By credit card directly with Anthropic.
4
Enter in .env
CLAUDE_API_KEY=sk-ant-... — enter your key in the bot's .env file.
5
Enable components
Trade check: ENABLE_CLAUDE_VALIDATION=true (master switch) + ENABLE_CLAUDE_TRADING_EXECUTOR=true
Exit check: ENABLE_CLAUDE_EXIT_VALIDATION=true (does not depend on the master switch). Everything is off by default.
6
Restart the bot
Done. Claude AI is now active. First test thoroughly in Paper trading mode.
# config/.env

# --- Claude AI API ---
CLAUDE_API_KEY=sk-ant-api03-...

# --- Master switch (trade check) ---
ENABLE_CLAUDE_VALIDATION=true

# --- Components ---
ENABLE_CLAUDE_TRADING_EXECUTOR=true
ENABLE_CLAUDE_EXIT_VALIDATION=true

# Restart bot...
[FuturesBot] Claude AI: 2/2 active
[FuturesBot] WebSearch: aktiviert (enabled)
[FuturesBot] Ready for trading.
Note: Note: the components can be activated individually. First check their decisions thoroughly in paper mode in the logs before adding further components.

Configuration in Detail

All parameters have default settings (not a recommendation); you switch on the components yourself. These options are available for adjusting.

.env — on/off switch

ENABLE_CLAUDE_VALIDATION Master switch for the trade check (the exit check depends only on ENABLE_CLAUDE_EXIT_VALIDATION)
ENABLE_CLAUDE_TRADING_EXECUTOR Trade validation before every new trade
ENABLE_CLAUDE_EXIT_VALIDATION Exit check on the time exit (can postpone the exit)

claude_config.py — Fine-tuning

CLAUDE_MIN_CONFIDENCE Minimum confidence for trade approval (50–80, default: 70)
CLAUDE_VALIDATION_TIMEOUT Max. wait time for the trade check (default: 30 s, with web search × 1,5); after that the signal is skipped
CLAUDE_EXIT_TIMEOUT Max. wait time for the exit check (default: 15 s, with web search × 1,5); after that, closing proceeds as normal
CLAUDE_EXIT_MIN_CONFIDENCE Minimum confidence above which a time exit is postponed (default: 70)
MAX_EXIT_DELAYS The maximum number of times a time exit is postponed (default: 3)
EXIT_DELAY_MINUTES Minutes until the next exit check (default: 3)
CLAUDE_MODEL_TRADE_VALIDATION Model for trade validation (default: claude-sonnet-4-5)
CLAUDE_MODEL_EXIT_VALIDATION Model for the exit check (default: claude-sonnet-4-5)

Security & Transparency

Every decision is recorded with its rationale in the log; the bot estimates the costs — your Anthropic billing is authoritative.

Clipboard
Everything is logged
Every API call with estimated token costs in claude_tokens.csv (excluding web search fees). Every decision with reasoning in the logs.
Shield
Block — no entries of its own
Claude AI, for trades, can block, but cannot trigger new trades. The exit check can also postpone a time exit (at most 3×). The AI is a checking layer, not an autonomous trader.
🔍
What the web search is for
In the trade check only for acute extreme events (Flash crashes, exchange outages, liquidation cascades). The exit check additionally evaluates news and sentiment of the last 1–2 hours.
🔒
Data stays private
Data only flows between you and Anthropic (USA). FuturesBot does not see any API responses, does not store any content and has no access to your Claude account. DSGVO (GDPR): You are the controller within the meaning of Art. 4 Nr. 7 DSGVO; transfer to the USA on the basis of Standard contractual clauses. Details: Privacy Policy § 17b · Anthropic Privacy.
⚡
Behavior on timeout
If Claude does not respond in time during the trade check, or an error occurs, the signal is skipped — the trade is not executed. In the exit check, the position is closed as normal in this case.
Money
Cost control
Claude is only consulted as the last check before the order, or after the pre-check for a time-based exit. The token log lists every call with estimated costs. You have full control over your Anthropic credit at all times.

AI-supported trading — 14 days free

The Claude AI integration is available during the trial period. You only need your own Anthropic API key and credit (at least $5 at Anthropic). The bot itself charges no additional costs for the AI function.

No API key yet? → Open Anthropic Console

The chat bubble emoji.

Questions? Our AI helps you!

Our AI support assistant is usually around the clock available (voluntary additional service with a daily limit) — right here on the page (💬 bottom right). It helps with setup, configuration, strategies, TradingView setup, server installation and more.

AI chat and FAQ answer many questions immediately. Email for everything else.

⚠️ Risk notice: FuturesBot is a technical analysis and automation tool — not a financial product, not investment advice, not an investment recommendation, not asset management. Claude AI is an additional checking layer and gives no guarantee of profits; AI assessments can be wrong. You set the strategies, leverage and use yourself. Trading leveraged crypto futures can lead to a total loss; stop-loss and protection functions do not reliably limit losses (price gaps, slippage, liquidation). Earlier or simulated results are no indication of future ones. Check every setting thoroughly in paper mode first. Full risk notice →
The chat bubble emoji. Questions? Ask me!
F
FuturesBot AI support
AI assistant · replies automatically