tu is an open-source terminal agent harness and automation utility maintained by Sahil Ahuja (sahil87). Built primarily with Node.js and integrating with fab-kit orchestration tools, it provides a command-line framework designed to coordinate automated tasks, scripts, and agentic workflows directly in developer environments. Operating as a CLI harness, tu structures tooling execution around extensible skill standards and modular components located within its Node.js runtime. It enables developers to execute discrete automation tasks, interface with project environments via directory-level configurations like direnv (.envrc), and execute scripted routines directly from the shell. Unlike standalone interactive coding agents, tu is structured as a lightweight execution harness that incorporates toolkit skill standards and fab-kit integrations. This design focuses on repeatable script execution, custom capability packaging, and environment-aware task runs rather than open-ended conversational code generation. tu is licensed under the open-source MIT License and is actively maintained on GitHub.
Node.js Harness Core — Implements execution logic and task management through a Node.js-based terminal architecture. - Toolkit Skill Standard Support — Integrates standardized skill interfaces for modular task execution and custom tool addition. - fab-kit Integration — Utilizes fab-kit components to manage build orchestration, task lifecycle, and script execution. - Environment Configuration — Supports automated shell environment loading through direnv (.envrc) integration. - Open-Source MIT Architecture — Freely available and customizable codebase hosted openly on GitHub.
Use cases
Use Case: Scripted Developer Workflow Automation Developers configure tu to execute standardized development and repository maintenance tasks through custom skill definitions. The harness manages execution context and task sequencing directly within terminal environments.
Use Case: Environment-Aware Task Execution Engineers leverage tu alongside direnv configurations to trigger project-specific build, test, and linting scripts with pre-loaded environment variables.
Use Case: Modular Agent Tool Integration Teams looking to adopt consistent agent tooling conventions can wrap custom automation routines into tu's skill structure for repeatable CLI execution.
Who it is for
tu is aimed at software engineers and DevOps developers who require a lightweight, Node.js-based harness to coordinate custom skills and terminal scripts. It is not intended for users seeking a turnkey, interactive conversational GUI assistant out of the box.
Install guide
Installation Clone the repository and install its dependencies using npm: bash git clone https://github.com/sahil87/tu.git cd tu npm install
First Run If you use direnv, allow the local .envrc configuration: bash direnv allow Build or link the package locally for CLI access: bash npm link
Example Session Run the CLI entry point to view available commands and skills: bash tu --help
Shell GPT — Shell GPT (sgpt) is an open-source command-line productivity tool created by Farkhod Sadykov (TheR1D). It connects standard terminal workflows directly…
Qwen Code — Qwen Code is an open-source terminal-based AI coding assistant maintained by the Qwen team (Alibaba Cloud / QwenLM). Built on…
Cline — Cline is an open-source autonomous coding agent runtime maintained by Cline Bot Inc. and an active open-source community. Available as…
Continue — Continue is an open-source AI code assistant and agent framework originally developed by the Continue team (continuedev) and recently acquired…
GitHub Copilot CLI — GitHub Copilot CLI is a terminal-based AI coding assistant developed by GitHub. It integrates the Copilot coding agent runtime directly…
Aider — Aider is an open-source terminal-based AI pair programming tool created by Paul Gauthier and the Aider-AI community. It operates directly…