What is Tau?

Tau is a coding agent that lives in your terminal. You type what you want in plain English — “explain this repo”, “add tests for the parser”, “fix this stack trace” — and Tau reads files, runs commands, and edits code to get it done, streaming its work as it goes.

It’s two things at once:

A tool you can use. A real terminal coding agent with an interactive UI, multiple model providers, durable sessions you can resume and branch, and a resource system for your own skills and prompts.
A project you can learn from. Tau is built to be read. Its source is organized into small, honest layers so you can see exactly how a coding agent works — from model streaming, to the agent loop, to tools and sessions.

What can it do?

Hold a conversation while reading, writing, and editing files and running shell commands in your project.
Work in an interactive TUI or as a one-shot command for scripts and pipes.
Talk to OpenAI, Anthropic, OpenAI Codex, OpenRouter, Hugging Face, or any OpenAI-compatible endpoint (including local models).
Remember every session, let you resume it later, and branch from any earlier point to explore a different path.
Stay usable in long sessions by compacting context automatically, and expose thinking modes on models that support them.
Extend itself with your own skills, prompt templates, and per-project instructions (AGENTS.md).

Who is it for?

You want a coding agent you can run and shape — install it, point it at a model, and work. Start with the Quickstart.
You want to understand how coding agents are built — read the core concepts, then the How Tau works section.

Where to go next

Quickstart — install Tau and run your first session in a few minutes.
Core concepts — the handful of ideas (agent loop, providers, tools, sessions, skills) that everything else builds on.
Guides — task-focused how-tos for the TUI, sessions, providers, and more.
Reference — exact CLI commands, slash commands, and keyboard shortcuts.