0xku/kon

Kon is a minimal coding agent (and also a highly opinionated one)

What it solves

Kon is a minimal coding agent harness designed to provide a lightweight alternative to bloated agent frameworks. It focuses on a tiny core prompt and a small, predictable toolset, allowing users to maintain a low token overhead while still having a powerful assistant for coding tasks.

How it works

Kon operates as a terminal-based agent that combines a minimal system prompt (under 270 tokens) with a set of core tools for file manipulation and shell execution. It externalizes project-specific context through AGENTS.md or CLAUDE.md files and allows for reusable behavior packs called "Skills" stored in markdown files. It supports a wide range of LLM providers via API keys or OAuth, including GitHub Copilot, OpenAI, DeepSeek, and local OpenAI-compatible endpoints.

Who it’s for

It is designed for developers who want a coding agent they can easily read, understand, and adapt without inheriting a large framework, as well as those who prefer a TUI (Terminal User Interface) for their development workflow.

Highlights

  • Minimalist Prompting: Keeps the fixed harness at approximately 1,000 tokens to save on costs and improve model focus.
  • Core Toolset: Includes built-in tools for reading, editing, writing, and searching files, as well as bash execution.
  • Interactive TUI: Features a rich terminal interface with fuzzy file searching (@), path completion, and slash commands.
  • Flexible Context: Supports project-level guidance via AGENTS.md and modular "Skills" for specialized instructions.
  • Session Management: Saves conversations as JSONL files, allowing users to resume, continue, or hand off focused contexts to new sessions.
  • Broad Provider Support: Integrates with multiple hosted providers and local models via OpenAI-compatible APIs.

Related

  • Project
    evotai/evot

    Evot is a lightweight terminal‑based coding agent that lets a LLM read, edit, write files and run shell commands. With a tiny prompt, four core tools, background execution, and plug‑in support for many LLM providers, it offers fast, low‑cost AI‑assisted programming directly from the command line.

  • Project
    Sentdex/minion

    A lightweight coding agent designed to eliminate context bloat, enabling efficient operation on local LLMs and remote APIs while providing tools to read, write, and execute shell commands.

  • Project
    SWE-agent/mini-swe-agent

    A minimal AI software engineering agent that uses only bash commands to solve coding tasks, providing a high-performance, hackable alternative to complex agent frameworks.

  • 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
    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.