Our MCP connector is public and operational: what we propose to clients, we use ourselves every day.
MCP connectors
AI assistants are becoming a channel: people ask, the assistant answers, and increasingly it acts. The Model Context Protocol (MCP) is the standard that lets Claude, ChatGPT and custom agents talk to your services: query your data, use your tools, start your processes. We do not preach it second-hand: Sequel is the first Italian consulting firm queryable by an AI assistant, through a public MCP connector. What we built for ourselves, we build for you.
The agentic channel today
1st
Italian consulting firm queryable by an AI assistant via MCP: us
2024
the year the MCP protocol was born, now adopted by the major AI platforms
24/7
your connector answers your customers' assistants, even when the office is closed
+42.8%
year-on-year growth of traffic to AI search: the channel where the connector works
MCP connectors: what we do
Connector design
What to expose, to whom, to obtain what.
The value lies in the choices, not in the protocol: which of your company's capabilities are worth making queryable (catalogue, availability, quotes, support), for which users and with what return. We start from the use case, not the technology.
Development and security
Authentication, permissions, rate limits, logging.
We build the MCP server with proper authentication, permissions and rate limits: the assistant sees only what it must see, does only what it is allowed to do, and every interaction is logged. Security is the project, not an option bolted onto it.
Publication and discovery
Registries, directories, on-domain discovery.
Registration in MCP registries, presence in assistant directories, discovery files on your domain: we make sure AI assistants find the connector and know how to present it to users.
Parallel surfaces: skills and GPTs
Same content, every platform.
The MCP connector is the core, but the agentic channel has more surfaces: skills for Claude, GPTs for ChatGPT, APIs for custom agents. We build the full ecosystem, coherent and maintainable from a single source of truth.
Measuring the channel
Queries, frequent questions, contacts generated.
Who queries the connector, what they ask, what converts: we instrument the agentic channel like any other business channel, so the investment is judged on numbers.
How we work on an MCP connector
01
Use case and scope
A workshop to identify what is worth exposing to assistants and with what goal: leads, sales, support, authority. The output is the connector's scope and its measure of success.
02
Development and testing
We build the MCP server, the authentication and the parallel surfaces, and test them with real assistants: the same questions your customer will ask, asked by us first.
03
Publication and watch
Registries, directories, on-domain discovery, usage monitoring. The channel launches measured; its evolution follows data, not fashion.
The case we know best: our own
Sequel, queryable by an AI assistant
Our MCP connector is public: an AI assistant can check a website's agent-readiness, know whether a Google update is rolling out, search our insights and request a quote, straight from the conversation. It is our technical calling card: what we propose to clients, we use ourselves first, every day.
4Tools exposed
2Surfaces (MCP + GPT)
24/7Operational
Let's talk
Want to be queryable by AI assistants?
Tell us about your company and what you would like an assistant to do for your customers. We reply within one business day.
Thank you. We'll be in touch shortly.
Frequently asked questions
MCP (Model Context Protocol) is the open standard, born in 2024, that lets AI assistants connect to external services: data, tools, actions. An MCP server is the connector that exposes your company's capabilities in this format: the assistant queries it during the conversation and answers with your information, not with approximate recollections.
Being present, in your own voice, in the channel where customers are moving their questions. Concrete examples: an assistant that checks availability and prices from your catalogue, prepares a quote, answers about shipping and procedures, books an appointment. Without a connector, the assistant answers about you using third-party sources; with one, you answer.
It depends on the scope: an informational connector (catalogue, content, contact requests) is a matter of weeks; a transactional one (orders, bookings, procedures) requires deeper integration with your systems. The initial workshop exists precisely to define scope, timing and investment before writing a line of code.
Yes, if the connector is designed well: you expose only what you decide to expose, with authentication and granular permissions, rate limits and full logging of interactions. It is the same principle as public APIs, plus the safeguards specific to the agentic context. Security design is an integral part of our work.
No. MCP was born at Anthropic but is an open standard adopted by the major platforms, ChatGPT included. And where a different surface is needed (a dedicated GPT, a skill, an API for custom agents) we build it in parallel, from the same source of truth.
Like any channel: queries received, most frequent questions, contacts and conversions generated, cost per interaction. We instrument the connector from day one, so that after three months the value discussion happens on numbers.
