About Agent Lsp
This project is an MCP server that gives AI coding agents access to language server intelligence. Agents often make incorrect edits because they cannot see who calls a function, what a rename will break or whether a build still passes. Language servers hold those answers, and this tool exposes them to MCP clients such as Claude Code.
It runs as a single Go binary that keeps a warm, stateful index of the workspace and routes each file to the right language server, for example gopls for Go, pyright for Python and the TypeScript language server for TypeScript. The documentation lists 66 tools, 30 languages verified in continuous integration against real language servers, and 24 agent workflows delivered as skills. It also offers speculative execution, which previews the diagnostic impact of an edit in memory before anything is written to disk, and phase enforcement, which blocks tool calls that violate the current step of a workflow.
It is aimed at developers who use AI agents on real codebases. The source is published on GitHub under the blackwell systems organization. The project does not publish pricing.