“LSP for AI” is the idea that AI coding tools need a shared standard for integrating with developer tools, by analogy with the Language Server Protocol (LSP). Before LSP, every code editor had to build its own support for every programming language; a language server written once now works with any editor that speaks LSP, turning an M×N integration problem into an M+N one. Addy Osmani’s agentic engineering glossary describes AI coding assistants as facing the same M×N problem today, each building its own ways to read code, run tests and search documentation, and presents a standard protocol through which any AI client can talk to any tool server as the equivalent fix.
Usage
The glossary treats protocols such as the Model Context Protocol as early steps toward this vision, and calls Anthropic’s MCP the most widely adopted AI tool protocol at the time of writing. The benefits it claims for engineers working with an AI Coding Agent are less lock-in to any one AI tool — swapping AI providers without rebuilding integrations — and better tooling overall, because a well-built server, for example an MCP server for GitHub or for a database, then serves every compatible client instead of a single one. It draws the parallel that LSP’s standardization led to an explosion of high-quality language tooling.
The glossary describes the vision as still early and evolving, and expects current protocols to iterate significantly as the field matures.