docs: Behat and AI

[carlos-granados]

Adds a documentation page explaining how Behat can be useful when using AI and which points to the new MCP server

[acoulton]

I'm not sure here whether you mean scenarios (e.g. .feature files) or the implementation of steps (e.g. in context files)?

But I think maybe we need a stronger note of caution that Behat (or any testing tool) only adds value if the tests are sufficiently robust to prove the behaviour the scenario is claiming.

Maybe something more like:

Suggested change

	Generated scenarios should be reviewed to ensure they accurately represent the intended behaviour.
	As with any specification or testing tool, generated scenarios and assertions should be reviewed carefully. It is critical to ensure that both the human-readable scenarios the executable step definitions accurately represent the intended behaviour. They also need to be robust enough to prove that the actual behaviour matches the expectations, even if the implementation changes.

[acoulton]

Thanks @carlos-granados this is helpful, and good to highlight the benefits Behat (and BDD) can bring to agentic development.

Just a couple of notes of caution from me, to make sure users are clear on the limits of what Behat itself (/Behat + and AI) will do for you "out of the box".

[acoulton]

Again here I think the claim is maybe a bit too strong: Behat won't protect you from an AI refactoring if your scenario & step definitions are too loose - or might add a lot of complication if they are too coupled to implementation details.

Maybe something more like:

Suggested change

	Behat ensures that behaviour remains stable as code changes. AI agents can safely modify or refactor code,
	With well-designed scenarios and step definitions, Behat ensures that behaviour remains stable as code changes. AI agents can safely modify or refactor code,

[carlos-granados]

@acoulton good suggestions, updated

[acoulton]

Thanks @carlos-granados