Versions:
agent-lsp is a stateful MCP (Model Context Protocol) server published by Blackwell Systems, currently available in version 0.19.2, with six versions released to date. Its purpose is to give AI agents reliable access to the Language Server Protocol by operating as a stateful layer over real language servers, rather than as a thin or stateless proxy. Where conventional integrations force agents to issue many low-level protocol requests and assemble the results themselves, agent-lsp keeps the underlying language server's semantic index warm, so that repeated queries against a codebase remain fast and consistent, and adds a skill layer on top that turns multi-step code operations into single, correct workflows. This design addresses a practical problem in agentic software development: language-aware tasks such as navigation, inspection, and refactoring normally require sequences of coordinated calls, and errors in any step can corrupt the whole operation; by encapsulating those sequences as skills, agent-lsp reduces both latency and failure rates. The software exposes 50 tools and supports 30 programming languages, making it broadly applicable across heterogeneous codebases, and it maintains persistent sessions so that agents retain context across interactions instead of re-initializing the language server for every request. Relevant categories for this catalog include developer tools, AI agent infrastructure, MCP servers, code intelligence, and language tooling. Typical use cases include enabling AI coding assistants to perform semantic code search, symbol navigation, type-aware analysis, and structured code edits; powering autonomous agents that must understand and modify large projects across many languages; and serving as shared language-intelligence infrastructure for teams standardizing on MCP-based agent workflows. Because it builds on real language servers, the semantic information agents receive reflects the same analysis that editors and IDEs use, which improves the correctness of agent-driven code operations. Organizations evaluating agent tooling can compare its six released versions to track the project's ongoing development toward the current 0.19.2 release.
Tags: