Chatbot MCP

Chatbot MCP – Intelligent Conversational Assistant with CPF Validation and Recharge Management

Chatbot MCP addresses the need for a lightweight, AI‑powered chatbot that can be seamlessly integrated into existing business workflows while handling domain‑specific logic such as CPF validation and balance manipulation. By exposing a Model Context Protocol (MCP) endpoint, the server lets AI assistants like Claude or Claude 2 invoke its capabilities in a structured way, eliminating the boilerplate code normally required to build conversational agents that interact with external data stores.

At its core, the server combines a GPT‑4 language model with a small in‑memory account ledger. Natural‑language requests are parsed by the LLM, which then delegates to typed handlers that perform CPF validation (accepting both formatted and unformatted numbers, checking length, verifying check digits, and rejecting trivial duplicates). Once a CPF is validated, the server can execute recharge or debit operations that adjust an account’s balance. The result of each operation is returned to the assistant in a concise, human‑readable format—e.g., “O saldo do colaborador com CPF 123.456.789‑09 é de R$ 50”—which the assistant can then present to end users.

Key features include:

• MCP API: The server implements the MCP specification, allowing any compliant AI client to discover and invoke its tools without custom adapters.
• Robust CPF Validation: Built‑in logic ensures only legitimate Brazilian tax identifiers are accepted, protecting downstream processes from malformed input.
• Dynamic Balance Management: Simple add/subtract operations on a virtual account ledger enable quick prototyping of financial workflows.
• CLI and REST Interfaces: Developers can test the service locally through a command‑line client or expose it over HTTP, making integration with existing pipelines straightforward.
• TypeScript & Zod: Strong typing and schema validation reduce runtime errors and improve developer ergonomics.

Typical use cases span internal corporate chatbots that manage employee benefits, customer support agents that handle prepaid balances, or any scenario where a conversational interface must perform controlled state changes on validated identifiers. By offloading the natural‑language understanding to GPT‑4 and encapsulating business logic behind MCP, developers can focus on higher‑level features such as authentication, persistence, or analytics without reinventing parsing logic. The result is a modular, testable chatbot stack that can be embedded in Slack, Teams, or custom web UIs with minimal friction.