Tau

Tau is an open-source terminal-based AI coding agent developed by Hugging Face. It provides an interactive CLI environment designed to assist developers with code generation, codebase exploration, and automated refactoring directly from the command line. In daily use, Tau operates through an interactive terminal interface supporting multi-turn dialogue, tool use, and file system manipulation. It accommodates multiple backend providers, supporting major cloud model APIs (such as Google Gemini with streaming output) as well as offline, local model inference via built-in llama.cpp integration accessible via the /local command. Tau differentiates itself through its tight integration with both open-source local inference stacks and cloud providers, alongside an extensible architecture. Developers can write extensions to inject tools or append custom sections directly into system prompts. It also offers fine-grained runtime control over reasoning models via CLI flags like thinking depth configuration. Tau is open source and hosted on GitHub under the Hugging Face organization. It is free to use, requiring developers to either supply their own API keys or run self-hosted local weights.

Tags: ai agent, cli, code assistant, coding agent, developer tools, OpenSource

Visit Tau

Key features

  • Built-in Local Inference — Execute models offline on local hardware using integrated llama.cpp support via the /local command. - Multi-Provider Support — Connect to various cloud inference providers, including robust streaming integration with Google models. - Custom Extension System — Author repository extensions that dynamically append system prompt sections and expose specialized tooling. - Configurable Thinking Budgets — Control model reasoning and thinking levels at startup using the --thinking (-t) command-line flag. - Native Python/uv Integration — Designed around modern Python packaging workflows with fast dependency management.

Use cases

Use Case: Offline and Private Local Development Developers working under strict privacy constraints can run code analysis and generation tasks completely air-gapped using local weights via the built-in llama.cpp runtime.

Use Case: Custom Agent Extensions for Repositories Teams can create workspace-specific extensions to inject project architecture rules, custom linters, and specialized prompts into the agent's context window.

Use Case: Interactive Terminal Debugging with Reasoning Models Engineers can execute multi-turn debugging sessions and modulate the model's reasoning effort directly via CLI flags to resolve complex algorithmic issues.

Who it is for

Tau is designed for developers, ML engineers, and CLI-centric programmers who require a customizable, open-source coding agent that supports both cloud APIs and offline local models. It is less suitable for users looking for a prepackaged GUI IDE plugin or a zero-configuration point-and-click tool.

Install guide

Installation Tau can be installed from source or managed via Python package managers like uv: bash # Clone the repository git clone https://github.com/huggingface/tau.git cd tau # Install dependencies and project environment using uv uv sync

First Run Set your desired provider API key in your environment before launching: bash # For cloud-backed models export GEMINI_API_KEY="your-gemini-api-key" # Launch the Tau CLI uv run tau To run locally with llama.cpp, launch Tau and use the /local command inside the session or point to a local model path.

Example Session Start an interactive session with reasoning mode enabled: bash uv run tau -t high "Analyze tests/test_parser.py, fix failing assertions, and run the test suite."

Part of CLI Coding Agents

Categories: AI Coding Agents, Code Assistants, Developer Tools

Related CLI agents

  • Plandex — Plandex is an open-source terminal-based AI coding agent designed to handle complex, multi-file engineering tasks in large codebases. Created by…
  • OpenCode — OpenCode is an open-source AI coding agent developed by Anomaly that operates across terminal environments, desktop applications, and IDE extensions.…
  • OpenHands — OpenHands (formerly OpenDevin) is an open-source autonomous AI software development agent and orchestration platform. It provides an extensible runtime environment…
  • Subtask — Subtask is an open-source CLI agent harness and orchestration tool developed by zippoxer in Go. It is built to coordinate…
  • EchoCoding — EchoCoding is an open-source agent harness and audio-augmented developer tooling layer created by launsion-boop. Designed to run alongside AI coding…
  • pi-reflect — pi-reflect is an open-source agent harness and self-improvement extension for pi coding agents, developed by jo-inc. It inspects agent session…
  • AI Tools
  • Categories
  • Industries
  • CLI Coding Agents
  • MCP Servers
  • MCP Categories