patys.dev

Dev-diary #6

Do we setup MCP servers? No. So we don’t want to add a lot of work around setting up anew microservice called MCP. But there are other things to consider: ways. MCP let’s you do it on scale.) Why? Agents have specific permissions, different access then people to different parts of the system. Adjusting MCP server will probably take the same amount of time to make them working in our internal system.

So we don’t want to add a lot of work around setting up a new microservice called MCP. But there are other things to consider:

how many agents will you use in future?
how many people will use their own agents in future?
does your company have access to proper tooling to use MCP?
We will have one team of agents working on multiple problems. -> We can say, we have one system
Due to security, policies etc., we are not using additional agents. -> We are building our agent, so let’s assume none as they will use our tool.
In theory we have. But practically, not the right people will use it. Mostly development team is using MCP and agents. -> So no real users of business related MCP servers.

But business people might want to use it. Yes, that’s why they will have their own agent (which we are building right now), so they don’t need to think about creating own tooling with MCP. They can just setup work in Slack and wait for results.

It’s dynamic world and technology, so my thought process can be wrong and outdated next month. But when to use it?

your team is using tools with MCP and AI access (e.g. VS Code or Claude Desktop),
you plan to use multiple, separated agents,
you plan let people use their own agents and just connect to MCP infrastructure.