The inspiration from @ncoop57.bsky.social
to get tmux to do the heavy lifting to make ShellSage work is genius! Thanks to this, ShellSage is <100 lines of code (plus comments and prompts).

But the result is amazing. Try it out and tell us how you go.

Details here: www.answer.ai/posts/2024-1...

OK. So your vision for MCP in editors is to work alongside the LSP server, and both could/should be enhanced by AI?

If so, I am missing an LSP++ (or MCP++) to support more UX than just autocompletion, e.g., next cursor position prediction, add typing info to the prompt, ...

🤞This would be a *huge* step forward to automate personal workflows in a safe way: calendar, mail access, ...

I did not hack enough with MCP :) Probably I am missing a lot of things.

With MCP I can see we can use a prompt spec.modelcontextprotocol.io/specificatio... with prefix/suffix arguments and so on. But AFAIK it is not described in the specification so I cannot "plug and play" supporting editors like in LSP.

The latter (LSP) let you to insert or replace in specific parts and access to the relevant context (workspace, typing info, ...).

Very interested to listen your take in how to implement a fill-the-gap experience like Copilot or "next cursor position" (aka Cursor magic tab) or access the workspace context, typing information, etc.

For example: spec.modelcontextprotocol.io/specificatio... vs microsoft.github.io/language-ser...

Maybe @juanlu.space knows a better way to do this with uv

Exactly :)

I hope it will get there. IMO At this moment it is not. It is a RPC for Agents, useful for chat-like experiences but not for coding.

Although venvs are path-specific you can install them in a generic folder (~/envs/) and activate them whenever you need them. Quick test: