justrach/codegraff

graff — a fast agentic coding harness in Zig: multi-provider, MCP, workflows, DGM evolution loop, TS/Python SDKs

What it solves

CodeGraff is a lightweight AI coding agent designed to handle development tasks directly from your terminal. It automates complex workflows—such as building applications, investigating performance issues, or processing data—by reading and editing files, executing commands, and using browser tools.

How it works

You provide a task in plain English, and the agent takes over. It uses sub-agents to perform parallel work, keeping context manageable by reusing setups and summarizing long transcripts. It supports various AI models (via API keys or browser-based logins) and can be extended with MCP (Model Context Protocol) servers to connect with external tools and data sources.

Who it’s for

It is for developers who want an efficient, terminal-based AI assistant that can handle multi-step coding tasks without heavy overhead. It is also suitable for those who prefer a desktop environment, as it integrates with the "Harness" app to provide a visual interface for chat, file review, and terminal access.

Highlights

  • Minimal footprint: A single 3.7 MB binary with zero runtime dependencies.
  • Parallel execution: Uses sub-agents to work on different parts of a task simultaneously.
  • Extensible: Supports MCP servers to add custom tools and capabilities.
  • Flexible integration: Can be used via CLI, as an external agent for editors like Zed, or through provided Python and TypeScript SDKs.
  • Cross-platform: Native support for macOS, Linux, and Windows.

Related

  • Project
    ad-naan/Adnify

    An AI-native engineering workspace that combines a professional code editor with governed multi-agent planning and direct execution capabilities.

  • Project
    AlgoNoRhythm/Flare

    A graph-first IDE for agentic coding that provides a live visual map of a codebase and a coordination layer for multiple AI agents to work together without colliding.

  • Project
    nurettincoban/ai-prd-workflow

    A structured, document-driven workflow for AI coding agents that enforces RFC-based development to ensure project consistency, traceability, and long-term planning.

  • Project
    robzilla1738/harness-terminal

    A native macOS terminal with GPU acceleration and session persistence that integrates with AI coding agents to notify users when they need attention.