{
  "id": 6327403,
  "title": "Does an LSP help a coding agent?",
  "url": "https://urgent.news/2026/09/08/does-an-lsp-help-a-coding-agent",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-09-08T20:50:46.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/scott_raisbeck_24ea5fbc1e/does-an-lsp-help-a-coding-agent-4a6f"
  },
  "original_language": "en",
  "account": "In the world of artificial intelligence, OpenAI has been making headlines recently with the introduction of a new class of models called GPT 6 - Astra. However, this article is not about that. Instead, it explores a different question related to coding agents and their performance.\n\nThe author, who is testing out an LLM (Language Model), discovered that their Pi agent lacked access to the Language Server Protocol (LSP). LSPs are software tools that IDEs (Integrated Development Environments) use to analyze code files with specific extensions. For example, if you're working with a .rs file and have a line of code that would cause a compiler error, your IDE can immediately highlight the error without you having to run the compiler.\n\nThe author experimented with adding an LSP extension to their Pi agent and found that it improved the agent's performance. They conducted evaluations using the Agent Shell's package and extension management feature, which allowed them to compare agents with different capabilities. The tests were run three times on three models: mimo-v2.5, muse-spark-1.3-contributor, and OpenAI's Luna, with the Pi agent having either no LSP or the @narumitw/pi-lsp@0.49.7 LSP extension added.\n\nThe results showed that the Pi agent with the LSP extension scored slightly higher than the agent without it, with an average of 81.3% compared to 81.0%. While this improvement was not drastic, it still represented a reduction in the number of tokens used by the models, indicating that the LSP extension helped to streamline the agent's operations. However, the author concludes that the benefits of using an LSP extension in their specific use case are inconclusive at this point and that they may leave the extension off unless they are pair programming with the LLM.",
  "summary": "It was a busy weekend with all the activity coming out of OpenAI. If you have been living under a rock, they introduced a new class of model, GPT 6 - Astra . I've been playing around with it somewhat. That is not what this article is about, though. I do not have an unlimited budget to run lots of evals against a model like this, so any opinion I have on it would be purely anecdotal. There are…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}