Claude Code LSP | Hacker News Reader

Claude Code LSP

karanbansal.in

77 pointsLexSiga39 comments

“ That's not an incremental improvement. That's a category change in how Claude Code navigates your code.”

Literally everything is written by AI now, even top HN articles.

I don't mind if people use AI to help them write, but when I see this kind of thing, it implies to me that they're barely even skimming it before posting. Surely people don't want this super cliche AI-hype-man tone in their blog posts, right? And if they haven't taken the time to at least skim through it and iterate on basic style, why should I assume it's worth my time to read it?

> Literally everything is written by AI now

Which explains the exceedingly stupid idea. If you want an LSP for Claude, you need an LSP for English. Wait, let my AI generate an article about that!

It's not as secret as they make it sound. Documented here: https://code.claude.com/docs/en/discover-plugins#code-intell...

Also the post is definitely AI written partially, but still useful I suppose.

I wish AI written submissions were banned.

Same. I can't this tone. I don't understand why modern LLMs can't show more variability in tone and writing style by default.

I wonder why circles can't have corners.

There's a "flag" button. See an AI-written submission? Press it.

I had great results using this for the past few months, it’s the same idea but implemented as an MCP: https://github.com/oraios/serena

Serena eats context for lunch (and dinner).

since its flagged - gotta comment; I am not the author of the post. I was reading it in passing and thought it was interesting enough to submit. Indeeed did not paid enough attention as to how much "ai written" it was.

It’s not enabled by default because there are still bugs eg race condition issues.

Check on GitHub.

For what's it's worth I never enabled anything and it proactively encouraged me to install various lsp plugins and prompted me to accept to install in most languages I've tried.

Had trouble with their two marketplace's as there's also another `anthropics-claude-code`.

Here's the commands to install the correct marketplace and LSP plugins:

/plugin marketplace add anthropics/claude-plugins-official

/plugin install typescript-lsp@claude-plugin-directory

See all LSP's names here: https://code.claude.com/docs/en/discover-plugins#code-intell...

In the same way that good AI coding requires testing, project management and architecture, good AI writing requires you to fill the editor role. Be ruthless. Read line by line. By all means tell the agent to fix stuff. If you don't do this, your blog posts sound generic and lazy.

It's not hidden at all, Claude pushes it even tho it poisons the context after every edit with false positives because it's always out of date. This feature should be hidden given how half baked it is.

Not sure if it's Claude Code or golsp, but I had to uninstall the plugin. Regularly makes Claude Code crash with golsp going to 100% CPU usage.

I assume if this was working it would allow Claude to much more quickly gain an initial understanding of a repo on initial use, no?

This blog post is in the uncanny valley of "looks and sounds nice, but a bit too nice, could be useful, could be useless AI-slop, idk".

I’m pretty sure it’s AI written. It has the common AI style of, “That’s not just X! It’s Y!”

Personally I find this annoying. I use AI for my writing but painstakingly try to maintain my own voice rather than lazily edit my prose into LinkedIn-speak.

I’m not usually bothered by this, but the style of this post made me feel mildly stressed while reading it. Not everything needs drama.

I built this myself for my agents a long time ago from ctags.

I hope I’m not talking to an AI. Can you explain how you wire up ctags to Claude code and keep them upto-date?

Just install universal-ctags from your package manager and it’s just a command line interface to build them / regen. Then you can build custom things on top of that. It outputs a big index to a cache dir. Not an AI for the record.

I'd test it but Claude is down

https://code.claude.com/docs/en/discover-plugins#code-intell...

Why are they doing this with a client plugin per server? It's antithetical to the whole point of LSP?

(At least they're reusing existing servers I suppose, but it stops me using whatever arbitrary one I want, as I could if there was just a single client with arbitrary configuration.)

I never got it working for even once. it's an epic failure on Anthropic

Works for all users except for one. Maybe it's not Anthropic's epic failure, but that one user's?

That is when the LSP works, now you've just introduced another thing that may fail.

For pytoy - maybe. Me personally, would love LSP-sourced completion in jetbrains. This dump bot suggests compile errors, APIs sucked out of thin air and other model-sourced nonsense. It does NOT read my code as I read and navigate code.

Id like LSP LLM in my IDE…

LSPs suck. Compilers should be able to re-use their cache for queries, interfaced with through a simple CLI.

They should not need more RAM than what the compiler uses.

Most LSPs don't work well for big projects or with non-standard setups like Bazel.

Edit: is this really so controversial it has to be downvoted?

But now your asking the compiler to also be a daemon. The compiler devs to add and maintain a not insignificant feature. The compiler to keep everything it caches for queries (whoops, look at that RAM usage climb!), and to cache data suitable for answering LSP queries (gosh! It's climbing a lot!)

Why does it need to be a daemon? Why isn't the on-disk cache enough? Running a compiler without making changes is instant. Many operations of LSPs feel slower than compiler's incremental compilation.

What cache is needed that the compiler doesn't already have?

Compilers only do one file at a time, but an LSP needs to parse all the files.

Also, an LSP needs to update incrementally when you edit the file, not just when you compile.

And not every language is compiled. And not everything that has an LSP is even a language. I don't think the gp comment was very thought through.

One of the many advantages of LSPs is that they can continue to work even under the condition of compilation error.

This is not unique to LSPs, a good compiler will do the same.

It's sad to see that the sane opinion is so heavily downvoted.

LSP as a protocol is fine, but the actual technical implementation of JSON RPC is braindead. Only web devs that don't know anything about native code could devise such an abomination. What happened to plugins and dll's?

From the intro section:

> That's not an incremental improvement. That's a category change in how Claude Code navigates your code.

I don’t know anything about the human(s) behind this project, assuming there are any, and intend no malice towards them, but when I encounter language like this, it just kills my enthusiasm for a project.

I wonder if that reaction is now, or will become, a majority one, and that AI flavoured language and products will face some audience headwinds if there aren’t indications of some level of human authorship / editing.

Reply on news.ycombinator.com