Ask HN: How should a team collaborate on their AI agent tooling?

Devs aren't used to having to agree on their stack. One person uses linux, vscode, and zsh; another uses a mac with neovim and fish. And that was totally fine.

But when you look at the bleeding edge agent-based stack, not only are there a lot of choices, it seems like there's more non-trivial configuration that you'd want to collaborate on as a team:

- You need lots of rules so your agents have a chance to zero-shot their tasks. But if someone uses Cursor and someone uses Claude Code, do you put the rules in .cursor/rules or in CLAUDE.md?

- Adding a rule is much more impactful than adding a single line of code. What does code review look like for these?

- For MCP servers and tools, is there some sort of effort to keep those in sync as well?

Of course, the AI vendors don't have an incentive to help with this kind of collaboration.

So, do devs who use the same tools end up in their own little conclaves? Or is it more common for someone to mandate that everyone on the same team shall use a given stack of AI tools?