Anthropic Tool Use (Function Calling) Integration for Your App

Anthropic Tool Use (Function Calling) Integration for Your App

Imagine a CRM manager spending 20–30 minutes gathering deal history, recent contacts, and open tasks before a meeting. With 50 managers, that's 1000 hours per month lost to routine. Tool Use from Anthropic turns Claude into an assistant that searches CRM, sends emails, and creates tasks in seconds — without risking SQL injection or exposing code. Claude doesn't execute code; it returns structured JSON with the tool name and arguments. Your application does the real work and returns the result. This is the foundation for agents working with live data.

Our engineers have over 5 years of commercial AI development and 50+ successful projects with RAG and Function Calling. We guarantee production-ready code with logging, retries, and monitoring. Contact us — we'll implement Tool Use in 3 days.

What Problems Does Tool Use Solve?

Without Tool Use, the model is limited to text generation. To get data from a database, you either give it direct SQL access (dangerous) or manually parse its responses and execute code. Tool Use solves both:

• Security: the model returns only JSON; the application controls execution.
• Reliability: JSON Schema defines argument formats, reducing errors.
• Performance: parallel tool calls cut agent response time by 2–3 times.

Additionally, Tool Use reduces model hallucinations with factual data — the model relies on real execution results, not its internal knowledge cutoff.

Basic Tool Use Loop

import anthropic
import json
from typing import Any

client = anthropic.Anthropic()

# Define tools
TOOLS = [
    {
        "name": "search_database",
        "description": "Search customers in CRM by parameters",
        "input_schema": {
            "type": "object",
            "properties": {
                "query": {"type": "string", "description": "Search query"},
                "limit": {"type": "integer", "description": "Number of results", "default": 10},
                "status": {
                    "type": "string",
                    "enum": ["active", "churned", "trial"],
                    "description": "Filter by status"
                }
            },
            "required": ["query"]
        }
    },
    {
        "name": "send_email",
        "description": "Send email to customer",
        "input_schema": {
            "type": "object",
            "properties": {
                "to": {"type": "string"},
                "subject": {"type": "string"},
                "body": {"type": "string"},
            },
            "required": ["to", "subject", "body"]
        }
    }
]

# Tool dispatcher
def execute_tool(tool_name: str, tool_input: dict) -> Any:
    if tool_name == "search_database":
        return search_crm(**tool_input)
    elif tool_name == "send_email":
        return send_email_via_smtp(**tool_input)
    raise ValueError(f"Unknown tool: {tool_name}")

def run_agent(user_message: str) -> str:
    """Full agentic loop with tool use"""
    messages = [{"role": "user", "content": user_message}]

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

        # No tool calls — return final answer
        if response.stop_reason == "end_turn":
            for block in response.content:
                if hasattr(block, "text"):
                    return block.text
            return ""

        # Process tool_use blocks
        tool_results = []
        for block in response.content:
            if block.type == "tool_use":
                print(f"Calling tool: {block.name}({block.input})")
                result = execute_tool(block.name, block.input)

                tool_results.append({
                    "type": "tool_result",
                    "tool_use_id": block.id,
                    "content": json.dumps(result, ensure_ascii=False),
                })

        messages.append({"role": "assistant", "content": response.content})
        messages.append({"role": "user", "content": tool_results})

Why Use Parallel Calls?

Executing multiple tools simultaneously speeds up the agent. When the model returns several tool_use blocks in one response, run them in parallel via asyncio. This reduces total time to the maximum among all tools. With three independent tools taking 0.5, 1.2, and 0.8 seconds, a synchronous approach takes 2.5 seconds, async takes 1.2 seconds — a 2–3x gain.

import asyncio

async def execute_tool_async(tool_name: str, tool_input: dict) -> tuple[str, Any]:
    """Asynchronous tool execution"""
    result = await asyncio.to_thread(execute_tool, tool_name, tool_input)
    return tool_name, result

async def run_agent_async(user_message: str) -> str:
    messages = [{"role": "user", "content": user_message}]

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

        if response.stop_reason == "end_turn":
            return next((b.text for b in response.content if hasattr(b, "text")), "")

        # Run all tools in parallel
        tool_use_blocks = [b for b in response.content if b.type == "tool_use"]

        results = await asyncio.gather(*[
            execute_tool_async(b.name, b.input)
            for b in tool_use_blocks
        ])

        tool_results = [
            {
                "type": "tool_result",
                "tool_use_id": block.id,
                "content": json.dumps(result, ensure_ascii=False),
            }
            for block, (_, result) in zip(tool_use_blocks, results)
        ]

        messages.append({"role": "assistant", "content": response.content})
        messages.append({"role": "user", "content": tool_results})

Synchronous vs. Async Loop Comparison

How to Force a Tool Call?

# tool_choice="required" — Claude must call at least one tool
# tool_choice={"type": "tool", "name": "..."} — call a specific tool

response = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=1024,
    tools=TOOLS,
    tool_choice={"type": "tool", "name": "search_database"},
    messages=[{"role": "user", "content": "Find customers with trial status"}],
)

Forcing a call is useful when you need to guarantee a critical action — for example, create a task or send a notification. The model can then return the call result without additional text.

How to Handle Tool Errors?

Errors are a natural part of production. Never end the dialogue on failure. Instead, pass the error description back to the model: it can choose another tool or ask for clarification. Step-by-step:

1. Wrap the tool call in try-except.
2. On success, add result to tool_results with type: "tool_result".
3. On exception, add a block with is_error: True and the error text.
4. Send tool_results to the model and continue the loop.

# Return errors to the model — it adapts behavior
try:
    result = execute_tool(block.name, 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,
        "content": f"Error: {str(e)}",
        "is_error": True,  # Model knows about the error and can adjust
    })

Example full response with error:

{
  "type": "tool_result",
  "tool_use_id": "toolu_abc123",
  "content": "Error: Connection timeout to database",
  "is_error": true
}

The Claude model, upon receiving such a result, can suggest retrying or using another tool.

Practical Case: CRM Assistant

From our practice: a client — a large retail chain with a sales department of 50 managers. Preparing for a client meeting took 20–30 minutes: collecting deal history, recent contacts, open tasks. We deployed an assistant on Claude with tools:

• get_customer_info — customer profile
• get_deal_history — deal history
• get_recent_activities — recent actions
• create_task — create a task
• get_calendar — free slots

Result: meeting preparation — 2 minutes via dialogue with the assistant. Time savings: 28 minutes per meeting. With an average manager salary of $30/hour, that's $14 saved per meeting, and with 10 meetings per day — $140 daily. Project payback — 3 months due to time savings from senior managers. Additionally, data entry errors dropped by 75% — tools work strictly by schemas.

What's Included in the Work

Timelines

• Basic tool use loop with 2-3 tools: 2–3 days
• Parallel calls + error handling: 1–2 days
• Production-ready with logging and retry: 1 week

Get a consultation on integration — we'll assess your project in 1 day and propose an optimal architecture. Our certified ML engineers have experience with OpenAI, Claude, Gemini, and open-source models. Write to us — let's discuss your scenario.

LLM Development: Fine-Tuning, RAG, Agents, and Production Deployment

Using GPT‑4 or Claude 3.5 Sonnet through a public API is not a solution — it's just a tool. When the requirement is to "make it like ChatGPT, but on our data," there is a real engineering challenge behind it: from prompt engineering to training a 70B model on your own infrastructure. End-to-end LLM solution development is a complex stack, and we have been doing it for over 5 years. During this time, we have completed over 20 projects in generative AI: from RAG systems for legal departments to custom support agents. Where exactly your task falls depends on data, latency requirements, budget, and how critical confidentiality is.

A typical situation: the client has already tried ChatGPT, but results are unstable — sometimes accurate, sometimes hallucinating. Or they need integration into a corporate portal while complying with security policies. Let's break down each layer of the stack in detail — from RAG to production deployment.

Why Do RAG Systems Break and How to Fix It?

RAG (Retrieval-Augmented Generation) looks simple: find relevant documents, put them in context, get an answer. In practice, it fails in several places.

Chunking without overlap. Classic mistake: chunk_size=512, overlap=0. If the answer lies across two chunks, retrieval won't find either with sufficient confidence. Solution: overlap 15–25% of chunk_size, or better yet, sentence-aware splitting with spaCy or NLTK instead of naive character splitting.

Poor embedder. text-embedding-ada-002 is good for general use, but on legal or medical texts, specialized models like E5-large-v2, BGE-M3, or fine-tuned sentence-transformers on domain data outperform it. Recall@5 differences can be 15–25%.

No re-ranking. Vector search optimizes for speed, not relevance. A cross-encoder re-ranker (ms-marco-MiniLM-L-6-v2, bge-reranker-large) after initial retrieval improves top-3 accuracy with acceptable latency (+50–150ms). This is often more impactful than improving the embedding model.

Hybrid search. Dense vectors alone work poorly on exact queries: names, SKUs, codes. BM25 (sparse) finds exact matches but misses semantics. Hybrid via RRF (Reciprocal Rank Fusion) is the optimal compromise. Qdrant, Weaviate, and pgvector 0.7+ support hybrid search natively.

When to Fine-Tune Instead of Prompt Engineering?

Prompt engineering solves ~70% of LLM adaptation tasks for a domain. The remaining 30% require fine-tuning. Three indicators: the model ignores a specific output format even with detailed prompting; the task requires deep knowledge of specialized vocabulary (medicine, law); you need to significantly reduce token costs by replacing a large model with a smaller specialized one.

LoRA and QLoRA are the standard for SFT. LoRA adds trainable low-rank matrices to attention layers. A typical configuration for Llama-3 8B: r=64, lora_alpha=128, target_modules=["q_proj","v_proj","k_proj","o_proj"] yields ~0.8% trainable parameters, training on one A100 40GB. QLoRA adds 4-bit quantization (NF4) and allows fine-tuning 70B models on two A100 40GB, though speed drops by half compared to bf16.

DPO instead of RLHF. Direct Preference Optimization requires only (chosen, rejected) pairs, not scalar reward signals. DPOTrainer from the trl library (Hugging Face) implements it in a few dozen lines.

Common mistake. A dataset of 500 examples, 5 epochs, validation loss 0.8 — seems fine. But on test, the model degrades on general instructions. Cause: catastrophic forgetting. Solution: add 10–20% general instruction-following examples (Alpaca, FLAN) to the training set to preserve original capabilities.

How to Choose a Base Model: 8B or 70B?

For most tasks, fine-tuning an 8B model is sufficient. 70B is needed when deep reasoning is required or the 8B baseline does not reach the required quality even after fine-tuning. Inference cost for Llama-3 8B via vLLM on A100 is efficient; the exact cost depends on volume.

What Does PagedAttention Bring to Production?

vLLM is the first choice for serving open-source models. PagedAttention is the key technical innovation: KV-cache is managed like virtual memory in an OS, without fragmentation. This yields 2–4x higher throughput compared to naive HuggingFace Transformers inference. The vLLM documentation confirms that continuous batching and PagedAttention are the standard for high-load LLM services.

Typical numbers on A100 80GB for Llama-3 8B (bf16): 400–600 req/s, P50 latency 200–400ms, P99 latency 600–900ms at concurrency 64. For 70B on two A100 with tensor parallelism: 80–120 req/s, P99 latency 1.5–2.5s. AWQ or GPTQ quantization reduces memory consumption by 2x with quality loss within 1–3%.

Multi-Agent Systems

Agents are LLMs with access to tools: search, code execution, API calls, database interaction. Common patterns:

• ReAct (Reason + Act): the model reasons → chooses a tool → observes the result → reasons again. LangChain and LlamaIndex implement it out of the box.
• Multi-agent orchestration: multiple specialized agents with a coordinator on top. Example: coordinator → researcher (search + summarization) → coder (code generation and execution) → critic (verification). Tools: AutoGen (Microsoft), CrewAI, custom implementation on LangGraph.

In production, agent systems are non-deterministic. Essential: guardrails, step limits, logging of each step, human-in-the-loop for critical actions.

How We Work: Stages, Timeline, Deliverables

What Is Included

We deliver:

• Technical documentation (model card, configs, deployment instructions)
• Access to infrastructure (code repository, trained weights)
• 1 month of post-deployment support (consultations, bug fixes)
• Customer team training (2–3 sessions on system operation)

Timeline: basic RAG prototype — 1–2 weeks. Fine-tuning with customer data — 3–6 weeks (including data preparation). Production system with monitoring and retraining — 2–4 months. Cost is calculated individually based on data volume, model complexity, and infrastructure requirements.

We guarantee the quality of the final model with performance benchmarks and ongoing monitoring. Our engineers have hands‑on experience with dozens of production LLM systems.

Want to evaluate your project? Leave a request — we will prepare a preliminary summary within 1–2 business days. Or get a consultation on choosing the approach: RAG, fine-tuning, or hybrid — we will tell you what works best for you. Contact us to discuss your LLM development needs. Schedule a free consultation today.