MyBenAI's built-in calculator and unit converter tools let your AI do math and convert measurements without external APIs or network calls. This is tool calling in action: the LLM requests a calculation, the local handler executes it, and the result comes back to the model—all on your phone, offline, with no round-trip to an API you pay for.
What Are Built-In Tools?
When you ask your AI a question that requires a tool, the model doesn't have to guess the answer. Instead, it recognizes the need, formally requests the tool with parameters, and then your phone executes the tool locally and returns the result.
For example: You ask "How many milliliters is 2.5 cups?" The model recognizes this is a unit conversion request. Instead of trying to recall a conversion factor (which it might get wrong), the model calls the unit converter tool with parameters cups: 2.5, target_unit: "ml". The tool runs locally on your device, returns 591.47 ml, and the model builds its answer from the real, accurate result.
MyBenAI ships with built-in tools for calculation, unit conversion, device status, contacts, calendar, clock, and dictionary. Many run locally without network calls. Currency exchange, weather, and crypto prices require live network data for current rates and conditions—but the requests originate from your device, not from OpenAI's servers. This is different from ChatGPT's plugin system, where every tool invocation goes over the internet.
How Tool Calling Works
Tool calling is a formal mechanism in language models. When a model is trained with tool-calling capabilities, it learns to emit a special structured output called a tool use request. This isn't just text; it's a machine-parseable instruction that says "execute this tool with these parameters."
The flow is simple:
- You ask the AI: "Convert 40 degrees Fahrenheit to Celsius."
- The model processes your question and determines that a temperature conversion tool is needed.
- The model emits a tool-use token sequence: tool_name: temperature_converter, params: {value: 40, from: "F", to: "C"}. This is not a guess; it's a deterministic output from the model's logits.
- The app's tool executor receives this request. It checks the tool allowlist (which tools are enabled), validates the parameters, and executes the local tool.
- The tool runs: 40°F = (40 - 32) × 5/9 = 4.44°C.
- The result, "4.44 C", is fed back to the model as context in the same conversation turn.
- The model reads the result and generates a natural-language answer: "40 Fahrenheit is 4.44 Celsius, which is very cold."
The entire cycle happens in milliseconds. No HTTP request, no API latency, no waiting for a remote server. Just local execution.
Calculator: Always Accurate Math
Language models are notoriously bad at math. A 7B model can fumble basic arithmetic; a GPT-4-class model is much better but still makes errors on complex calculations. This is because LLMs compute probabilities of the next token, not deterministic math.
With a built-in calculator tool, your AI never guesses at arithmetic. You ask "What's 17% of $2,500?" The model recognizes this needs calculation and calls the calculator tool. The tool computes: 2500 × 0.17 = 425. The model then answers with certainty: "That's $425."
This is faster and more reliable than asking the LLM to reason through the math. The model still provides context and explanation, but the computation is offloaded to a deterministic tool.
Without this, you get answers like "Let me think... 17% of 2500 is approximately 420 to 430, probably closer to 425" (the model hedges because it isn't sure). With the calculator tool, you get: "$425, exactly."
Unit Conversion: Instant, Offline, Accurate
Unit conversion is similarly powerful when automated. MyBenAI's unit converter handles length, weight, volume, temperature, time, data storage, speed, energy, and more. Examples:
- Distance: miles to kilometers, nautical miles to meters, light-years to AU.
- Weight: pounds to kilograms, ounces to grams, stones to tons.
- Volume: cups to milliliters, gallons to liters, tablespoons to teaspoons.
- Temperature: Fahrenheit to Celsius to Kelvin, with proper offset handling.
- Data: gigabytes to terabytes, megabits per second to gigabytes per hour.
- Energy: kilowatt-hours to joules, BTU to calories.
When you ask "I bought 3 kilograms of coffee. How many pounds is that?" the model calls the converter: kilograms: 3, target_unit: "lb". Response: 6.61 pounds. Instant, no rounding errors, no cloud dependency.
This is crucial for international users, scientists, engineers, and anyone working with measurements across unit systems. The accuracy and speed cannot be matched by an LLM trying to recall conversion factors.
Why Local Tools Beat ChatGPT Plugins
ChatGPT's plugin system is powerful in concept but has real drawbacks. When you ask ChatGPT a question that requires a plugin, ChatGPT decides which plugin to call and then makes an HTTP request to that plugin's endpoint. This introduces several costs:
- Latency: Each tool call requires a network round-trip. On a typical internet connection, that's 100–500ms per request. Complex queries might invoke multiple tools sequentially, adding seconds to the response time.
- API costs: Each plugin call can cost money. A third-party calculation API or unit converter might charge per request. Over thousands of daily requests, this adds up.
- Reliability: If the remote service is down or rate-limited, the plugin fails. Your AI assistant is now only as reliable as third-party infrastructure.
- Privacy: Every tool invocation sends data to a remote server, which logs the request and can analyze it.
With MyBenAI's local tools, all of these problems disappear. Calculation and conversion are instant, free (no API calls), always available (the code runs on your phone), and private (the numbers you are working with stay on your device and never reach a third-party API service).
Tool Allowlists and Privacy
You might worry: "If the model can call tools, can it call any tool?" The answer is no. MyBenAI enforces a tool allowlist. Only specific tools can be called, and you control which ones are enabled.
In Settings, you can see all available tools and toggle them on or off. You can also see a tool access log showing when tools were called and with what parameters. This transparency means you always know when and how your AI is using tools.
For MCP (Model Context Protocol) connectors and REST integrations, there's an additional layer: per-connector tool allowlists. If you add a home automation connector, you can restrict it to temperature and lights only, blocking access to locks and cameras.
Examples of Tool Calling in Action
Scenario 1: Budget planning - You're planning a trip and want to convert your savings to different currencies. You ask "I have $5,000 saved. What's that in EUR, GBP, and JPY at today's rates?" The model calls the currency tool three times, fetches current rates, and compares the results for you. The rate lookup needs a connection — live rates have to come from somewhere — but the arithmetic on top of them runs on your phone, and you are not paying an AI vendor per request to do it.
Scenario 2: Recipe scaling - You have a recipe in cups and teaspoons but your scale measures grams and milliliters. You ask the AI to convert the entire ingredient list. The model calls the unit converter multiple times (cup to ml, teaspoon to ml, etc.) and returns a full gram-and-milliliter version of your recipe. Done in seconds.
Scenario 3: Scientific calculation - You need to compute a formula with several steps: "If I invest $10,000 at 7% annual return, compounded monthly, for 15 years, how much will I have?" The model uses the calculator tool to work through the compound interest formula, step by step, and returns the exact result (approximately $27,897). Each step is accurate because the math is delegated to a deterministic tool, not approximated by the LLM.
The Honest Trade-Off
Local tools are fast and private, but they're limited to what's built in. If you need a specialized calculator or a niche unit system, MyBenAI's tools won't cover it. You're also limited by the scope of built-in tools: you can't add custom tools without modifying the app.
Cloud-based AI services like ChatGPT have the opposite trade-off: they offer a vast plugin ecosystem and custom integrations, but at the cost of latency, API fees, privacy, and reliability. For everyday tasks—math, unit conversion, weather, currency—local tools are strictly better. For specialized or rarely-used tools, a plugin ecosystem is more flexible.
Function Calling Across Your AI Workflow
Tools aren't just for calculations. MyBenAI uses function calling for everything: checking the weather, querying your calendar, reading device status, searching Wikipedia, and more. Each tool is local, each executes instantly, and each keeps your data private.
If you want to explore how tool calling is orchestrated across agents and multi-step workflows, read Function Calling & Tools: How Offline AI Gets Things Done. For a broader view of how to use MyBenAI without internet, see Offline AI Assistant: Use AI With No Internet.
Beyond calculator and unit converter, MyBenAI also supports Weather, Currency & Crypto Prices as reference tools. And if you're curious about building your own function-calling workflows without API keys, Function Calling Without API Keys: Free Local Tools dives deeper.
Built-in tools transform your AI from a general-knowledge engine into a precise, capable assistant for tasks that require accuracy. A calculator that never miscalculates, a unit converter that never hesitates, all running locally and instantly—this is what offline tool calling enables. Ready to experience it? Download MyBenAI today and start using tools that just work, with zero network latency and zero API costs.