From Software Engineer to AI Engineer - Part 5: Scaling your tool belt
We learned about tools and wrote them ourselves. This is cute, but an application writing all of its own tools is not scalable. In software development, we put functionality in libraries and frameworks and reuse it across projects. MCP (the Model Context Protocol) is all about exposing tool catalogs to models. An MCP server is maintained either by a SaaS provider or a company's internal AI…
In software development, engineers typically build and reuse libraries and frameworks to increase scalability. However, when it comes to AI applications, creating all tools from scratch can be impractical. This is where the Model Context Protocol (MCP) comes in. An MCP server maintains a catalog of tools that AI applications can use, much like a library of pre-built tools for backend engineers. MCP servers are usually hosted by SaaS providers or internal AI platform teams.
An MCP server publishes a catalog of tools with names, typed parameters, and descriptions. An AI application retrieves this catalog using an MCP client, which is then passed to a model for processing. Communication between the client and server occurs via the MCP protocol, which is based on the JSON-RPC 2.0 protocol. The model never communicates directly with MCP; instead, tools within the catalog handle the requests.
To illustrate, let's consider PayIQ's tools: calculate_refund_cost and search_payments_knowledge_base. To publish these tools over MCP, a developer creates an app/mcp_server.py file. The file initializes a FastMCP object with the payiq-tools catalog, including the two tools. To start the server, the developer runs python -m app.mcp_server, and then uses curl to initialize the MCP session and retrieve the catalog.
The session ID is obtained and then used to fetch the list of tools. This catalog serves as a tool repository for the AI application, enabling it to efficiently utilize the available tools without the need for each application to develop them independently.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
