Stock & inventory assistant
Owners and customers ask about stock in plain language — quantities on hand, totals, percentages, what's running low — and every number in the answer is computed by a deterministic math engine, not the model. Zero hallucinated digits, even at scale. That exactness is the moat: general-purpose chatbots guess at numbers, MathBot never does.
General-purpose LLMs routinely get simple number comparisons and decimal addition wrong. MathBot prevents that by design.
Numbers are computed with exact decimals and fractions, never binary floating-point rounding errors like 0.1 + 0.2 ≠ 0.3.
Claude and OpenAI behind the same interface — pick a provider per request, no vendor lock-in.
One endpoint, one API key. Send a question in plain language, get back an exact structured answer.
Number comparisons, arithmetic, percentages, rounding, number properties (even/odd/prime), sorting.
The LLM is only ever allowed to understand the question — never to invent a number.
The user asks something like "what is 15% of 240" or "which is bigger, 3/7 or 0.43".
The model only ever calls a structured tool and extracts the numbers from the text — it never computes anything itself.
A deterministic Python engine (Decimal / sympy) computes the precise result and returns it in the answer.
Production API keys are issued per license — contact us to get one.
# curl example curl -X POST https://your-domain.com/api/v1/ask \ -H "X-API-Key: mb_your_key" \ -H "Content-Type: application/json" \ -d '{"question": "Is 17 a prime number?", "provider": "anthropic"}' # response { "ok": true, "provider": "anthropic", "operation": "number_properties", "text": "17 is odd, prime. Divisors: 1, 17." }