Tool Use & Search/Retrieval

{
  "name": "get_weather",           // Snake_case function name — clear and specific
  "description": "Get the current weather conditions and forecast for a city. Use this when the user asks about weather, temperature, or forecast for any location. Returns current conditions and a 3-day forecast.",
  "input_schema": {
    "type": "object",
    "properties": {
      "city": {
        "type": "string",
        "description": "The city name, e.g. 'London', 'Tokyo'. Include country code for ambiguous cities: 'Springfield, IL, US'"
      },
      "units": {
        "type": "string",
        "enum": ["celsius", "fahrenheit"],
        "description": "Temperature units. Default to celsius unless user specifies."
      }
    },
    "required": ["city"]
  }
}

TOOLS = [
    {
        "name": "web_search",
        "description": "Search the web for current information, news, facts, and data. Use when the user asks about recent events, specific facts, or anything that may have changed after your training cutoff. Returns a list of relevant excerpts with source URLs.",
        "input_schema": {
            "type": "object",
            "properties": {
                "query": {"type": "string", "description": "Search query — 2-6 words, specific"},
                "num_results": {"type": "integer", "description": "Number of results, 1-10. Default 5."}
            },
            "required": ["query"]
        }
    },
    {
        "name": "calculator",
        "description": "Perform precise mathematical calculations including arithmetic, statistics, and financial formulas. Use this for any calculation that requires precision — do not calculate in your head. Returns the numeric result.",
        "input_schema": {
            "type": "object",
            "properties": {
                "expression": {
                    "type": "string",
                    "description": "A valid mathematical expression as a Python expression string. E.g.: '(1500000 * 0.08) / 12' or 'sum([12, 15, 8, 22]) / 4'"
                }
            },
            "required": ["expression"]
        }
    },
    {
        "name": "query_database",
        "description": "Query the company's internal database for customer data, orders, product inventory, and metrics. Use when the user asks about specific accounts, orders, or internal data. Returns rows matching the query.",
        "input_schema": {
            "type": "object",
            "properties": {
                "table": {"type": "string", "enum": ["customers", "orders", "products", "metrics"]},
                "filters": {"type": "object", "description": "Key-value pairs to filter records, e.g. {\"customer_id\": \"C1234\"}"},
                "limit": {"type": "integer", "description": "Max records to return. Default 10, max 100."}
            },
            "required": ["table"]
        }
    }
]

import anthropic
import json

client = anthropic.Anthropic()

# Your actual tool implementations
def web_search(query: str, num_results: int = 5) -> list:
    """Call your search API here."""
    # ... implementation
    return [{"title": "...", "snippet": "...", "url": "..."}]

def calculator(expression: str) -> float:
    """Safe expression evaluator."""
    import ast, math
    return eval(expression, {"__builtins__": {}, "math": math})

TOOL_IMPLEMENTATIONS = {
    "web_search": web_search,
    "calculator": calculator,
}

def run_agent(user_message: str) -> str:
    """Run the tool use loop until Claude is done."""
    messages = [{"role": "user", "content": user_message}]

    while True:
        response = client.messages.create(
            model="claude-opus-4-5",
            max_tokens=4096,
            tools=TOOLS,
            messages=messages
        )

        # Add Claude's response to the conversation
        messages.append({"role": "assistant", "content": response.content})

        # If Claude is done (no more tool calls), return the final text
        if response.stop_reason == "end_turn":
            for block in response.content:
                if hasattr(block, 'text'):
                    return block.text

        # Process all tool calls in this response
        tool_results = []
        for block in response.content:
            if block.type == "tool_use":
                tool_fn = TOOL_IMPLEMENTATIONS.get(block.name)
                if tool_fn:
                    try:
                        result = tool_fn(**block.input)
                        tool_results.append({
                            "type": "tool_result",
                            "tool_use_id": block.id,
                            "content": json.dumps(result)
                        })
                    except Exception as e:
                        tool_results.append({
                            "type": "tool_result",
                            "tool_use_id": block.id,
                            "is_error": True,
                            "content": str(e)
                        })

        # Return tool results to Claude for the next iteration
        if tool_results:
            messages.append({"role": "user", "content": tool_results})

# Usage
answer = run_agent("What's the current EUR/USD exchange rate and how has it changed this week?")
print(answer)

{
  "name": "search_knowledge_base",
  "description": "Search our internal knowledge base for product documentation, support articles, policy documents, and internal guides. Use this tool BEFORE answering any question about our products, policies, or procedures. Always search before responding — do not rely on memory. Returns relevant document excerpts with source document names.",
  "input_schema": {
    "type": "object",
    "properties": {
      "query": {
        "type": "string",
        "description": "The search query. Rephrase the user's question as a direct search query. E.g. if user asks 'can I get a refund?' search for 'refund policy'"
      },
      "max_results": {
        "type": "integer",
        "description": "Maximum number of document chunks to retrieve. Default 3, max 8."
      }
    },
    "required": ["query"]
  }
}

You are a research assistant with web search capabilities.

PROCESS FOR RESEARCH QUESTIONS:
1. Identify what you need to find: break the question into searchable sub-questions
2. Search for each sub-question using the web_search tool
3. Evaluate the search results: note the source quality and recency
4. Synthesize the information into a coherent, cited answer

CITATION FORMAT:
For every factual claim, include the source in brackets: [Source: Publication Name, URL]
If you searched but couldn't find reliable information, say so explicitly.

QUALITY STANDARDS:
- Prefer primary sources (official sites, academic papers, original reporting)
- Flag outdated information (more than 2 years old for rapidly-changing topics)
- Distinguish confirmed facts from interpretations or estimates
- If sources conflict, present both views and note the discrepancy

NEVER answer factual questions from memory alone when web_search is available.
Always search first, then synthesize from real results.