MCP server. I wonder what would be real life (testing) use cases

Discussions
, , ,
1

I am reading a bit more about MCP servers now and trying to understand how they work under the hood.

While I am playing with Playwright MCP, replicating some simple weather forecast MCP server, will learn more about Postman API MCP server, I am immediately wondering what testing use cases, quality use cases, any use cases would be to build its own custom one.

Maybe I am looking for far ahead with own custom MCP server, but the idea of practical application is my next question. What MCP servers (any) can help us with? Curious to hear your thoughts, ideas, etc

3 Likes
2

I just commented on another post on whether playwright mcp could be used from exploratory testing.

“hands free interactive guided UI testing” might be interesting, a bit like pair testing with the llm but with the mcp at the keyboard.

I have not got as far though as customising my own mcp, I just do not know enough to see that potential yet.

2 Likes
3

After seeing @angiejones talking about goose (codename goose | codename goose) and MCP mashups, I am still trying to find the time to try that out

2 Likes
4

I have been playing with Debbie O’Brien’s tutorials and still figuring out the real use case to apply or whether it’s just another tool amongst many out there. https://debbie.codes/

5

@nat,

MCP servers remain somewhat fresh in the testing realm; however, I can already envision some interesting applications for it, especially when utilized alongside Playwright or Postman.

From a testing perspective, an MCP server could be used as a bridge between natural language instructions and automated execution. Instead of writing a lot of boilerplate code, you describe the test in plain English, and the MCP understands it and triggers Playwright or API test scripts.

Things I observe in real-life scenarios:

Fast prototyping of tests – Instantly create functional or E2E tests without hand-coding the framework.

API regression checks – A Postman MCP could run endpoint validations on demand or on a schedule.

Custom quality dashboards – Your MCP server might gather test results from various tools and environments for instant reporting.

Ad-hoc exploratory automation – Testers should be able to execute rapid, one-off automation commands in natural language without touching the codebase.

Creating a custom MCP server might feel too much “ahead,” but it can prove to be a game changer when it comes to scaling tests in teams where not every tester is a seasoned coder. It’s not to replace engineers, but rather to streamline and automate testing so that there are fewer errors being made and it is accessible and fast.

If you’re curious about how MCP servers can be applied in real-world testing, especially for zero-code E2E automation, this video is a great reference:

E2E Testing WITHOUT Writing Much of a Code! :scream: | MCP Server + Playwright
By ExecuteAutomation → Watch here

It walks through:

  • Setting up an MCP server for test automation

  • Creating full end-to-end tests without manual coding

  • Real website testing examples

  • Playwright + MCP integration

  • Running natural language test instructions

  • Building professional workflows in minutes

A solid starting point if you’re exploring practical MCP testing use cases.

2 Likes
6

  • Testing an MCP Playwright server. The only use case is a junior tester with no programming experience. But this makes no sense at all.

  • Exploratory testing with MCP Playwright server: slow as hell, and what are the advantages?

  • Writing automated tests with the help of natural language and MCP Playwright: Not workingn or maintenable code

Whenever you see or hear “vibe coding”, run away :slight_smile:

7

Juan,

These are interesting point and I think its worth getting a few more views on this.

MCP playwright server. Only use case is junior tester. My coding is not great but I tend to only value light coverage at UI layer and on the face of it this seems to do that.

Examples: “Find all contacts on website and verify names and emails are accurate and displayed in correct format”, “Evaluate the site from an accessibility aspect”.

It can quickly do that with structured tests. Yes its fairly basic, will it scale and will it do more complex things I do not know yet but for light coverage for a junior tester, its not bad.

What I’ve not seen though is comparisons with and without the mcp, would mcp only make sense if no access to code, if access to code are tests generated by llm alone better quality as they can also add test ID’s and maybe create cleaner POM files for use. This I’d like some views on.

As an exploratory tool, maybe not playwright but I quite like the idea of full on voice testing, pairing with an llm as I sit back and drink my coffee, “find suitable test credit card numbers with different expected responses, run them through the payment steps and provide an analysis of the results include both displayed messages and response codes”, that analysis then guides my next steps but the idea that its Guided interactive testing. Is this sort of thing still way off?

The slowness is a factor for now, could the mcp/llm combo do those actions quicker than I hands on, not so sure, but I still like the idea of it.