samchon/ttsc

What it solves

ttsc addresses the fragmentation and performance overhead of the TypeScript toolchain. It replaces the standard tsc to integrate type-checking, linting, formatting, and custom compiler transformations into a single pass, reducing the need for multiple separate tools and CI jobs. It also solves the problem of "hallucinating" or inefficient AI coding agents by providing them with a structured compiler knowledge graph instead of raw file text.

How it works

It acts as a drop-in replacement for tsc, reading the same tsconfig.json and emitting the same JavaScript. It hooks into the TypeScript compilation process to run plugins that can perform type-driven code generation or validation. It includes a specialized MCP (Model Context Protocol) server that exposes the resolved type-checker graph to AI agents, allowing them to query the codebase's structure without reading every file.

Who it’s for

TypeScript developers who want a more integrated build pipeline, teams requiring strict requirement coverage (via the evidence graph), and developers using AI coding agents who want to improve agent accuracy and reduce token costs.

Highlights

• Unified Toolchain: Combines building, type-checking, linting, and formatting into one stream.
• Type-Safe Execution: ttsx ensures a project fully type-checks before any code is executed.
• Evidence Graph: Turns documentation and requirements into compile-time obligations that fail the build if not acknowledged in code.
• Compiler Knowledge Graph: An MCP server that provides AI agents with structured compiler data, significantly reducing token usage.
• Plugin Ecosystem: Supports compiler-powered libraries like typia for automatic runtime validator generation.