The idea that AI coding tools need a shared standard - just like LSP unified code editors.
Before LSP (Language Server Protocol), every code editor had to build its own support for every programming language. VS Code needed its own TypeScript support, Vim needed its own, Emacs needed its own. LSP changed that - a language server written once works with any editor that speaks LSP. The M×N problem became M+N.
"LSP for AI" is the idea that we need the same thing for AI coding tools. Right now, every AI coding assistant builds its own integrations with dev tools - its own way to read code, its own way to run tests, its own way to search documentation. This is the same M×N problem LSP solved, just for a different era.
Protocols like MCP are early steps toward this vision. Instead of each AI tool building custom integrations, a standardized protocol lets any AI client talk to any tool server.
Standardization accelerates everything. When LSP made it easy for any editor to support any language, we got an explosion of high-quality language tooling. The same thing is starting to happen with AI tool protocols.
For agentic engineers, this means less lock-in to specific AI tools. If your workflow tooling speaks a standard protocol, you can swap between AI providers without rebuilding all your integrations. It also means better tools overall - a well-built MCP server for your database benefits every AI tool in the ecosystem, not just one.