TypeScript-Go toolchain

ttsc

Build with TypeScript-Go. Run scripts only after they type-check. Report lint rules, unmet requirements, and plugin findings in the one diagnostic stream your CI already gates on.

10x

faster type checks

TypeScript-Go checks natively, in parallel. JavaScript tsc does neither.

800x

faster lint loop

Rules reuse the AST and types the compiler already built. ESLint parses it all again.

100%

requirement coverage

Every configured requirement is acknowledged by name, or the build stops. Counted per obligation, never pooled.

90%

fewer agent tokens

Agents read one compiler-resolved graph over MCP instead of crawling source files.

Read the guide->Try the Playground

compiler path

ttsc

build / check / watch

ttsx

checked script runner

@ttsc/lint

rules as compiler diagnostics

@ttsc/evidence

requirements as compiler diagnostics

@ttsc/graph

code graph for coding agents, over MCP

plugins

AST + Checker before emit

src/index.ts:3:7 - error TS2322

Type errors and lint rules land in the same diagnostic stream.

[Toolchain]

One compiler path, not a pile of wrappers.

Keep the existing tsconfig.json. The same project graph feeds the CLI, runner, linter, editor, bundlers, and plugins.

read tsconfig

type-check

run plugins

emit, execute, or report

ttsc

Compiler

Build, check, watch, emit, and run source transforms on top of TypeScript-Go.

build / check / watch

ttsx

Runner

Execute a TypeScript entrypoint only after the owning project passes type-check.

typed tsx-style scripts

@ttsc/lint

Lint and format

Report rules as TS diagnostics, apply autofixes, and format through the same compiler pass.

fix / format / TSxxxxx

@ttsc/evidence

Requirement coverage

Make every configured requirement demand a named acknowledgement from the code, test, or document that satisfies it.

spec as compile error

@ttsc/graph

Coding agents

Hand a coding agent a checker-resolved graph of your codebase over MCP, so it stops grepping and re-reading files.

code knowledge graph

plugins

Compiler-powered libraries

Run transforms and checks on the types the checker already resolved, on every build and every ttsx run.

typia / nestia / your own

[Diagnostics]

Type errors and lint errors should look like one failure.

ttsc can print rule violations as TS diagnostics. Local runs, CI, and the editor all point at the same file, line, and rule.

type errors

lint violations

format fixes

editor underlines

$ npx ttsc --noEmit

src/index.ts:3:7 - error TS2322: Type 'number' is not assignable to type 'string'.
src/index.ts:2:5 - error TS17397: [prefer-const] Use const instead of let.
src/index.ts:1:1 - error TS11966: [no-var] Unexpected var, use let or const instead.
1  var x: number = 3;
   ~~~~~~~~~~~~~~~~~~
2  let y: number = 4;
       ~~~~~~~~~~~~~
3  const z: string = 5;
         ~
4  
5  console.log(x + y + z);
Found 3 errors in the same file.

[Requirement coverage]

Your spec, as a compile error no agent can skip.

@ttsc/evidence fails the build once per obligation, because one reference never covers another. @evidence <target> <reason> closes one, by naming a unit of the spec and why this declaration answers for it.

An AI coding agent has to clear those errors to finish. Coverage reaches 100% on its own, as the residue of the ones it closed.

Wire one claim Read the guide

CouponStackingNotice.tsx

/**
 * @evidence docs/discount.md#coupon-stacking <reason...>
 * @evidence POST:/orders/{orderId}/coupons   <reason...>
 * @evidence {@link hooks.useCouponStacking}  <reason...>
 */
export function CouponStackingNotice(props: IProps)

$ npx ttsc

error TS16411: [evidence/graph] Missing acknowledgement for
  'docs/discount.md#coupon-stacking'
error TS16411: [evidence/graph] Missing acknowledgement for
  'POST:/orders/{orderId}/coupons'
error TS16411: [evidence/graph] Missing acknowledgement for
  'useCouponStacking'
Found 3 errors.

codex gpt-5.6-luna, four subjects. Thirteen edges and token spend: full benchmark.

[Coding agents]

Your agent should ask the compiler, not grep.

@ttsc/graph hands a coding agent a checker-resolved graph of your project, over MCP. On the agent-cost benchmark, Claude agents answer reading zero files, cutting tokens by roughly 90%.

Connect an agent Full benchmark

.mcp.json

{
  "mcpServers": {
    "ttsc-graph": {
      "command": "npx",
      "args": ["-y", "@ttsc/graph"]
    }
  }
}

The whole setup. It never writes into CLAUDE.md or AGENTS.md.

GPT-5.6 Sol, shared onboarding question. Every model and repository: full benchmark.

[Plugins]

Plugins get the compiler's eyes.

A plugin is not a string replacement pass. It runs beside the TypeScript-Go AST and Checker, then reports diagnostics or rewrites source before JavaScript is emitted.

Explore plugins

runtime code from types

typia

Validators, JSON serializers, LLM tools, and Protobuf codecs generated before runtime.

backend contracts

nestia

NestJS routes, OpenAPI documents, SDKs, and E2E scaffolding backed by typia.

First-party utility plugins

@ttsc/banner

Attach a package JSDoc banner to emitted files.

@ttsc/paths

Rewrite path aliases into relative JS and declaration imports.

@ttsc/strip

Remove debug calls and debugger statements before emit.

[Playground]

Try the compiler path in the browser.

The playground boots the WebAssembly build of ttsc in a worker. Your source stays in the tab while you test diagnostics, transforms, and emitted output.

Open the Playground->Read wasm docs

01edit TypeScript

02run ttsc.wasm

03inspect diagnostics

04compare output

globalThis.ttsc.build()

Same engine shape as the CLI, packaged for in-tab demos and plugin playgrounds.

[Sponsors]

Built with help from people who keep the work moving.

Sponsor support funds the quiet work behind ttsc: TypeScript-Go upgrades, platform binaries, plugin compatibility, documentation, and the playground.

Sponsor ttsc

Thank you for your support