Skip to content

Event app: AICD - Mumbai

Event page

30 minutesTechnologyIntermediateEnglish

Rethinking Testing for MCP Servers

to

Room 1

Speakers

About this session

MCP(Model Context Protocol) servers introduced a new kind of client: one driven by an LLM (Large Language Model). Unlike the clients developers have worked with so far, this one is non-deterministic. It can call tools in unexpected sequences, with unpredictable inputs, and its main source of truth is its interpretation of the tool description.

We can't reliably test how an LLM will call our tools, and we definitely cannot have such tests in a CI pipeline. This makes it clear that the MCP server is the only component under our control, and it must be tested thoroughly. If the server is well-designed, the chances of the LLM doing the right thing goes up. A well-designed server is well-tested.

While reviewing several popular MCP servers, I found that most define “works correctly” as checking only happy paths and doing a basic check that tool descriptions exist. A quick online search also suggested there are few resources that offer guidance on testing MCP servers. It felt like the mental model does not exist yet, or is still emerging.

This talk introduces a practical testing model for MCP servers that builds on existing testing fundamentals, but adapts them to a non-deterministic client. Attendees will learn how to treat tool descriptions as functional contracts in their tests, how to design tests that cover real-world and deliberate out-of-order tool call sequences, emulating what an LLM would attempt, and how to validate error channels so that, when things go wrong, the server returns errors that are actually useful for an LLM to recover or respond appropriately. These are practical techniques that can be applied immediately by testers and developers.

The goal is to showcase a testing model that can keep up with the fast-moving MCP ecosystem and keep it resilient.