Managing MCP server maintenance at scale, how do you handle API breaking changes across multiple clients ?

[M4shmalow]

Pre-submission Checklist

• I have verified this would not be more appropriate as a feature request in a specific repository
• I have searched existing discussions to avoid duplicates

Your Idea

Hi everyone,

I'm researching a pain point ikeep hearing from team deploying MCP-based agents in production, and i'd love to get feedback from people actually building in this space.
The problem i'm observing is when a thrid-party API changes (Stripe, Github, Saleforce, Notion..), every MCP server that connect to that API needs to be updated. If you're managing agent for multiple clients, this means :
1 breaking change upstream x N clients impacted x M agents per client = dozens or hundreds of manual fixes to deploy

And the detection is often reactive you find out when a client complains, not before.
I'm working at the feasibility of a centralized MCP gateway layer that would sit between AI agent and MCP servers: handling auth, routing, observability, and most importantly centralized maintenance. Fix once in the gateway, all clients get the update instantly.

I doesnt want to pitch anything just want to know if this solved problem, a know pain point or something teams have just accepted as part of the job.
Any feedback, war stories or thoughts on the approch would be incredibly valuable. Thanks in advance 🙏

Scope

• Protocol Specification
• SDK Features
• Documentation
• Developer Experience
• Other

[miqui]

@M4shmalow - great post. This concern also applies when API is in a deprecation+sunset part of its lifecycle. I would imagine during deprecation that the MCP server would start communicating some warning.