# AI integration and MCP server development

> Useful AI for your business: assistants that answer from your data, automatic document extraction, agents, and MCP servers to use your system from Claude or ChatGPT, with metered costs.

Page: https://app.azanello.com/en/services/ai-integration-mcp

Adding AI to a small business is not "installing ChatGPT": it means connecting a language model to your data and processes so it does one concrete job (answering questions, sorting orders, reading receipts, drafting quotes) with limits and human control. An MCP server (Model Context Protocol) lets assistants like Claude or ChatGPT use your system safely. It is quoted as a fixed price or in phases, and AI usage is metered and shown in your portal. The assistant that quotes projects on this site is an example of what I build.

## Who it is for
- Teams receiving lots of emails, documents or photos to sort and enter.
- Businesses that want an internal assistant answering from their own information.
- Companies with their own system that want to run it from Claude or ChatGPT.
- Anyone who tried AI and doesn't know how to make it work every day.

## What it solves
- Slow answers because information lives in a thousand places.
- Manual data entry from PDFs, photos or emails.
- Fear of the AI making things up, or of an unpredictable monthly bill.
- AI experiments that never got past a demo.

## What's included
- One use case, one metric: What the AI does and how it is measured, defined before building.
- Assistant on your data: Searches your documents, catalog or database and cites where each answer comes from.
- Document extraction: Pulls data from PDFs, photos and emails into a clean format.
- MCP server: Your system available to Claude or ChatGPT, with permissions, two-step confirmation for writes and an audit log.
- The right model: Claude, GPT or Gemini, chosen by cost and quality for your case.
- Costs under control: Spending caps and usage visible in your portal.

## How I work
1. Use case and metric: Which task, which data and what "good" means (for example, "27 of 30 questions answered correctly").
2. Prototype on real data: Evaluated with real cases before investing in the final version.
3. Production with limits: Permissions, spending caps and human confirmation for anything that writes, charges or sends.
4. Operation: Usage and results in plain sight; tuning through the maintenance plan.

## Options
- Focused AI automation or prototype — To try AI on one concrete task before investing more. — Fixed price: 50% to start and 50% on delivery. — 1 to 3 weeks — https://app.azanello.com/cotizar/asistente?lang=en&servicio=ai&opcion=prototipo
- Integrated assistant or agent, or an MCP server for your system — So AI works with your business data, with permissions and logging. — Phased or setup + monthly, depending on whether it keeps running. — 3 to 8 weeks — https://app.azanello.com/cotizar/asistente?lang=en&servicio=ai&opcion=asistente
AI usage is billed as metered with an agreed cap: no surprises on the invoice. Tap "Ask about this" on the option closest to what you need: the assistant asks a couple of questions and gives you an estimated range for your case in a few minutes.

## Frequently asked questions
Q: What is an MCP server and what is it for?
A: MCP (Model Context Protocol) is an open standard, created by Anthropic and adopted by the major AI assistants, that lets an assistant use external tools and data. With an MCP server for your system you can ask Claude or ChatGPT things like "which invoices are due this week?", with permissions and a log of every action.

Q: How much does AI cost per month?
A: It depends on volume and model. It is metered: you see tokens and cost in your portal, and a cap is set so there are no surprises.

Q: Is my data used to train the AI?
A: It uses the providers' business APIs, which under their current policies don't train models on data sent through the API. On top of that, only the information each task needs is sent.

Q: What if the AI gets something wrong?
A: That is why there are acceptance criteria and real-case testing before launch. Anything that writes data, charges money or messages a customer goes through human confirmation.

## How to get a quote

Describe the idea in plain words to the AI assistant: https://app.azanello.com/cotizar/asistente?lang=en&servicio=ai. It asks a few questions and shows an estimated range after email verification; Alejandro reviews every request and sends a written proposal with a fixed price. No account needed.

## Who's behind it

Alejandro Zanello is a freelance software developer based in La Falda, Córdoba (Argentina). He builds websites, custom web apps and platforms, CRMs, AI WhatsApp bots, automations and integrations for small businesses in the US, Mexico and Latin America. He works remotely, with written proposals, pay-per-delivery and a portal where every client sees their tasks, hours and payments.

Contact: me@azanello.com · https://azanello.com · https://www.linkedin.com/in/alejandrozanello/ · https://github.com/AleSZanello
