May I ask what your workflow actually looks like? There's been a fair bit of clojureposting over the last few days, and I've decided to jump in and learn.

I love the idea of an AI integrated repl (like what Jeremy Howard and team have done with solveit), it's far more in line with my preferred vision of the AI augmented future of coding. Less "swarm of agent" more, "learn with the agent".

Sure!

Setup is:

- nvim --listen /tmp/nvim — starts Neovim with a socket Claude can connect to

- /mcp in Claude Code — enables the Neovim MCP server, gives Claude direct control of Neovim

- lein repl in the nvim terminal

- Claude reads .nrepl-port, runs :ConjureConnect — REPL is live!

The loop is so dope:

- Claude writes code directly into my .clj files

- Then evals it into the running process via Conjure

- Sees the result in the REPL, iterates if wrong, all in the same conversation turn

- wrap-reload middleware means the web server hot-reloads changed namespaces on the next request

Hmmmm… seeing this workflow makes me wonder if I can do this with Ruby (the integration between agent and repl)

Absolutely, spin up a TCP server in Ruby, make it run `eval` on what it receives and tell the agent how to call it with netcat or whatever, works like a charm :)

> I've decided to jump in and learn.

https://clojure.org/news/2026/03/30/zero_to_repl

Also I love your take.

Feel like a holy grail to me to back in the hot seat with repl driven but I can drop in and figure things out... all on the JVM. Madness!