AI Trading Agent — Real Data, Not Hallucinations

AI Summary Response

The /v1/summary endpoint returns a pre-computed AI bias score and narrative — ready to pass directly into your LLM's system context.

{
  "success": true,
  "symbol": "EURUSD",
  "timeframe": "H1",
  "bias": "bullish",
  "confidence": 82,
  "narrative": "EURUSD broke 20-EMA with RSI divergence confirming momentum. MACD histogram expanding.",
  "indicators_analyzed": 42,
  "key_levels": {
    "support":    [1.0820, 1.0805],
    "resistance": [1.0860, 1.0875]
  }
}

Full Agent Pattern

Tool-use loop: the LLM decides which TickAtlas endpoint to call, you execute the HTTP request, feed the JSON back as a tool_result, repeat until the agent has enough context to answer.

import anthropic, requests

client = anthropic.Anthropic()
KEY = "YOUR_API_KEY"

TOOLS = [{
  "name": "get_summary",
  "description": "Live AI market bias + narrative for a symbol/timeframe.",
  "input_schema": {
    "type": "object",
    "properties": {
      "symbol":    {"type": "string"},
      "timeframe": {"type": "string"},
    },
    "required": ["symbol", "timeframe"],
  },
}]

def run_tool(name, args):
    if name == "get_summary":
        return requests.get(
            "https://tickatlas.com/v1/summary",
            headers={"X-API-Key": KEY},
            params=args,
        ).json()
    return {"error": "unknown tool"}

msgs = [{"role": "user", "content": "Analyze XAUUSD on H4."}]
while True:
    r = client.messages.create(
        model="claude-sonnet-4-6",
        max_tokens=1024,
        tools=TOOLS,
        messages=msgs,
    )
    if r.stop_reason != "tool_use":
        print(r.content[0].text)
        break
    block = next(b for b in r.content if b.type == "tool_use")
    result = run_tool(block.name, block.input)
    msgs += [
        {"role": "assistant", "content": r.content},
        {"role": "user", "content": [{
            "type": "tool_result",
            "tool_use_id": block.id,
            "content": str(result),
        }]},
    ]