Description
A course to build MCP servers that give AI agents access to your systems (a database, the ticketing tool, the ERP) without opening the whole door. MCP (Model Context Protocol) is the open standard that assistants such as Claude, GitHub Copilot or ChatGPT, and your own agents, use to call external tools. On the course you write a server, give it permissions and test it until it can be handed to users.
Course details
| SIXE code: SXIA09 | Category: Artificial Intelligence / Agents |
| Delivery: Online and on-site | Duration: 2 days |
Who it is for
Developers and architects who need to connect an assistant or an agent to internal applications, and security teams who have to approve that connection.
Prerequisites
Python or TypeScript, REST APIs and JSON. Having used an assistant with tools helps, but no previous MCP experience is needed.
What you will learn
- How MCP works: tools, resources and prompts, and the two transports: local over stdio and remote over HTTP.
- Writing a server: with the official Python or TypeScript SDK, exposing an internal system through well-described, narrowly scoped tools.
- Authentication and permissions: OAuth for remote servers, read-only by default and human approval before any write.
- Connecting it to an agent: from a desktop assistant and from your own agent with LangGraph and a local model.
- Testing and monitoring: MCP Inspector, logging every call, and prompt injection arriving through what the tools return.
You work with a sample internal application and finish with your own MCP server, connected to an agent, with its permissions and audit log.
What you take away
- An MCP server exposing an internal system with narrow permissions.
- The configuration to connect it to an assistant and to your own agent.
- A security checklist for approving new MCP servers.
To design the agent from scratch, the AI agents for companies course comes before this one. And if you would like us to integrate it for you, that is our secure AI agents service.
Frequently asked questions
How long is the course?
2 days.
Is it online or on-site?
Both: at your offices or remotely, with a live instructor.
What do I need to know beforehand?
Python or TypeScript, REST APIs and JSON. Having used an assistant with tools helps, but no previous MCP experience is needed.
Can we use the public MCP servers that already exist?
Yes, and the course shows how to assess them before installing: what permissions they ask for, who maintains them and what data they see.
Does it work with local models?
Yes. MCP does not depend on the model: the same server works for a commercial assistant and for an agent running an open model on your own servers.
Can it be tailored to my team?
Yes. We adapt the examples and labs to your team’s tools and processes, and it can be combined with other AI courses.
