A small foundation.

lspkit is a set of Rust crates for building language servers. Your analysis runs once, in one engine, and both an LSP server and an MCP server read from it.

SharpLsp · C# and F# language server for VS Code ↗SharpLsp profiling a real .NET app in VS Code. Watch the walkthrough ↗

One engine

EngineApi is the boundary between your analysis and the protocols. Every report carries a generation number, so a client always knows which state of the workspace produced an answer.

Two surfaces

The LSP server and the MCP adapter consume the same engine. Protocol handling stays at the edges; analysis lives in one place.

Shared plumbing

File tracking, configuration, sidecar processes, tree-sitter parsing, transport and profiling are ready-made. Use the parts that fit your server.

CheckoutPipeline.cs with inline heat bars: ParseCatalog 66.8 percent, PriceBasket 25.1 percent, RenderInvoice 7.6 percent — **SharpLsp** · C# and F# language server for VS Code ↗Hot lines, painted in the source.

busy.osp in VS Code with inline heat bars: collatz 76.4 percent, longest 73.0 percent, fib 22.5 percent — **Osprey** · native functional language with a VS Code extension ↗Heat on Osprey source.

Next steps

Getting started: add the crates and implement EngineApi.
Crates: what each crate and feature provides.
Profiler: CPU and memory profiling for any language server.
Examples: SharpLsp and Osprey running the profiler in VS Code.
The architectural field note: why one engine serves two surfaces.