# Turing

A Claude Code-style AI programming assistant available as both a command-line CLI and an Electron desktop app, sharing one core agent loop that edits files, runs shell commands and manages background tasks.

Turing is an AI programming assistant similar to Claude Code, offered as a command-line CLI and an Electron desktop application built on an npm workspaces monorepo, with the core agent logic extracted into a shared, reusable package used by both front ends. A user asks questions or gives instructions in natural language, and the agent responds and edits code files directly.

Its capabilities include reading, writing and glob-searching files, running shell commands with permission interception, compressing and persisting context across long conversations, an agent loop that continues until the model naturally stops rather than a fixed turn limit, scheduled cron tasks, task management with dependency chains, Git worktree management for parallel branches, a pluggable skill system, an internal multi-agent message bus, and external channels for Telegram, Feishu and WeChat Work. A structured trace log records tool calls for performance analysis.

It connects to any model through a messages-style API, configurable per provider including a bearer-token auth mode. It is written in TypeScript.

Website: https://github.com/Turing-Neo/Turing-coding-agent