AbdoKnbGit/tau

Your adaptive AI agent for your work.

What it solves

Tau is an adaptive coding agent designed to reduce the cost and complexity of AI-assisted development. It solves the problem of "noisy" context windows and high token costs by using optimized tools, a persistent Python kernel for complex data tasks, and a system that minimizes unnecessary data transfer between the model and the user.

How it works

Tau acts as a harness that connects to 28 different AI providers via native adapters, avoiding proxy layers. It implements a full agent loop including tools, skills, and subagents. Key technical mechanisms include:

  • Persistent Python Kernel: Executes complex workflows (like analyzing multiple CSVs) in a single turn rather than multiple iterative turns.
  • Optimized File Handling: Uses ripgrep for fast search and reads files by "signal" (skeleton-first) to avoid loading entire large files into context.
  • Shadow Git Repo: Maintains snapshots of the working tree for "time travel," allowing users to restore previous states without affecting the project's actual git history.
  • Native Browser Automation: Controls a real Chrome window to inspect frontends and perform web tasks.
  • Subagent Management: Spawns named, persistent subagents that can work in parallel and coordinate file edits to avoid overlaps.

Who it’s for

Developers who want a high-performance coding agent that supports a wide variety of LLM providers, offers granular control over costs (via "cheap mode"), and provides advanced project management features like session snapshots and remote collaboration.

Highlights

  • Native Multi-provider Support: Direct API adapters for 28 providers with automatic tool-schema sanitization.
  • Time Travel: Save, diff, and restore project snapshots via a hidden shadow repository.
  • Remote Control: Share sessions via local Wi-Fi or Cloudflare tunnels for mobile access or team collaboration.
  • Self-Learning: Proposes reusable lessons after tasks to build a persistent memory across projects.
  • Terminal Diagrams: Renders architecture and flow diagrams directly in the terminal.
  • Broad Rule Compatibility: Automatically reads rules from Cursor, Copilot, Cline, and Windsurf configurations.

Related

  • Project
    huggingface/tau

    A readable, terminal-based coding agent that can edit code and run commands, designed as both a tool and a teaching example for building agentic systems.

  • Project
    tontinton/maki

    Maki is a terminal‑based AI coding assistant that minimizes LLM token usage through tools like a tree‑sitter index, sandboxed code execution, and sub‑agent task delegation. It supports dozens of model providers, offers a fast Rust TUI, Lua‑based plugins, and safety features such as permission‑aware command parsing and SSRF protection.

  • Project
    0xku/kon

    A minimal coding agent harness with a tiny core prompt and a small built-in toolset, designed for lightweight and adaptable AI-assisted development.

  • Project
    pulseaiclub/phi

    A high-performance, terminal-based AI coding assistant that provides a lightweight, keyboard-driven interface for managing files, running shell commands, and executing complex coding tasks with LLMs.

  • Project
    qybaihe/mu

    A coding agent featuring a judgment kernel that uses small, fast models to handle non-coding decisions, reducing token cost and latency while optimizing context management.