ThinkLangType-safe AI for TypeScript

Structured LLM outputs, agentic tool calling, and multi-provider support. Use as a JS/TS library or a standalone language.

Get Started (Library)

Get Started (Language)

GitHub

Works in Any JS/TS Project

import { think, agent, zodSchema } from 'thinklang' — use the runtime directly in any JavaScript or TypeScript project. Zero config, full type safety.

Type-Safe Output

Define structured types with Zod schemas or .tl type declarations. The AI is constrained to return valid, schema-conforming data.

Agents & Tools

Define tools and run multi-turn agentic loops as library calls or language constructs. The agent calls tools until it produces a final answer.

Multi-Provider

Model-agnostic: Anthropic, OpenAI, Google Gemini, and Ollama out of the box. Swap providers with a single environment variable.

Production Ready

Guards for output validation, Confident for uncertainty tracking, cost tracking, built-in test framework with snapshot replay, and an LSP-powered IDE.

AI as Syntax

think, infer, and reason are keywords in the ThinkLang language. The compiler validates types, catches errors, and generates optimized LLM calls.