I started out thinking one global config was fine — just put the skills and MCP servers everyone needs at the user level and call it done. But over time I was wearing two hats: a frontend developer at my day job, and a fullstack developer on side projects. The configs started bleeding into each other.

The specific frustrations that pushed me over the edge: - Project-level settings had to be redone for every new repo - Marketplace and plugin-based configs needed manual toggling depending on which context I was in - Trying an experimental setup, then cleaning it up afterward, was tedious every time

Beyond config mess, the bigger realization was about personas. As I started doing more AI agent development, my developer identity kept splitting. More roles, more personas — and a single global Claude Code environment can't cleanly represent all of them.

That's when I thought: nvm and pyenv let you switch environments by profile. Why not Claude Code? So I built clenv.

clenv manages multiple Claude Code profiles. Each profile is an isolated ~/.claude directory (CLAUDE.md, MCP servers, hooks, agents, skills) backed by its own git repository. Free and open source (MIT).

  clenv init                            # backs up ~/.claude, creates default profile
  clenv profile create work --use       # create + switch instantly
  clenv profile create agent-prod --from agent-dev  # clone from existing

  clenv commit -m "add GitHub MCP server"
  clenv diff HEAD~1..HEAD
  clenv log --oneline
  clenv revert abc123f
  clenv tag v1.0 -m "production agent config"

  clenv profile export team-standard -o team.clenvprofile
  clenv profile import team.clenvprofile --use

Per-directory auto-switching works like .nvmrc:

  clenv rc set work    # pin profile to this directory
  clenv rc show

  brew tap Imchaemin/clenv && brew install clenv
  cargo install clenv

Would especially love feedback from people doing AI agent development — that's the use case where environment isolation feels most important.