Comment by cookiengineer
6 days ago
Seeing everything these days being about vibe coding, I feel a little old with my VIM setup and my LSP servers who I already thought were a nice productivity increase.
The problems I have with the stuff relating to MCP is that the tech around it is developing so fast that it's hard for outsiders to catch up with what the best working setup is, for example.
What would you do, for example, if you want to selfhost this?
- which models (qwen ai coder?)
- which api (with ollama? Bolt? Aider? Etc)
- how to integrate PRs with a local gitlab/gogs/forgejo instance? Do you need another MCP agent for git that does that?
- which hardware dependencies to run it?
I am currently trying to figure out how to implement a practical workflow for this. So far I'm using still a synchronous MCP agent setup where it basically runs on another machine in the network because I have a too unperformant laptop to work with.
But how would I get to the point of async MCP agents that can work on multiple things in my Go codebases in parallel? With the mentioned PR workflows so that I can modify/edit/rework before the merges?
The author makes a lot of claims and talks always about that their opponents in the argument are not talking about the same thing. But what exactly is the same thing, which is reproducible locally for everyone?
I haven't even gotten around the integrating LSPs into my vim workflow -- I feel positively ancient.
I can recommend govim, it's using LSP behind the scenes. I'd also recommend tagbar because it uses the newer ctags fork which supports much more languages.
[1] https://github.com/govim/govim
[2] https://github.com/preservim/tagbar
Thanks. I'll have a more thorough look tomorrow. I'm glad to see that neither seem to depend on NeoVim, which is also something I haven't bothered to take up.
2 replies →
I have https://github.com/prabirshrestha/vim-lsp in my minimal config and pretty much the default configuration for it. I only use LSPs for completion, auto imports, rename, and goto definition (less useful when the project is small) and that has been enough for me.
The self hosted and vim / nvim / emacs integrated llm / agentic / MCP situation is being implemented real time as we speak. There's no good canonical solution right now, especially for self hosting. If you are willing to make API calls to openapi or Microsoft, your options expand dramatically.
I think most people are using off the shelf paid solutions like cursor, windsurf, warp terminal, Claude code.
I'd love to get a set up with a locally hosted full version of deepseek.
I use vim as well, and one editor-agnostic coding assistant that is pretty decent is Aider. It is a little buggy sometimes, but it's been a blast to experiment with. I run it in a separate split with Tmux. Its whole schtick is to use it as a pair programming device. So I'll edit code in one split and use Aider in another, alternating at key points. This has worked pretty well for personal projects. I typically don't do this at $dayjob, though.
Why do you want to selfhost it? I just use https://github.com/Exafunction/windsurf.vim in NeoVim like a super-autocomplete.
Multiple reasons:
- Germany, so I have crappy and unreliable internet (read as: 300kB/s or less pretty often due to overbooked cable internet connections).
- Can't share my code with the US (as a European) because I work in cyber defense.
- I want a reproducible work setup, and something that will survive 2 years and/or the startup exit of the proposed tools.
There is no way to build a setup that will last two years. Hell, this new crop of agentic coding hasn't even been around for 4 months.
You should add Claude code to your setup and be amazed. Tmux with vim, Claude code and git work trees is amazing - now the choke point is… me.
I was not aware you could self-host Claude Code.
[flagged]