dox012/nano-claude

claude code simplified (~2000 Lines)

What it solves

nano-claude provides a lightweight, educational alternative to complex AI coding assistants. By stripping away the massive codebase of enterprise-grade tools, it offers a transparent, readable implementation of an agentic coding assistant that developers can study, modify, and run locally.

How it works

The project is built as an incremental series of versions, each adding specific capabilities like tool-use loops, markdown rendering, session persistence, and sub-agent orchestration. It functions as a command-line interface (CLI) that connects to Anthropic's API, allowing the AI to interact with your local environment through a set of seven core tools, including file editing, shell command execution, and content searching.

Who it’s for

This project is designed for developers who want to understand how AI coding agents are built from the ground up. It is ideal for those who prefer a minimalist, hackable tool over "black box" applications and want to learn about agentic architecture through a step-by-step, version-controlled roadmap.

Highlights

  • Minimalist codebase: ~2,000 lines of TypeScript across 19 files.
  • Educational roadmap: Includes a learning guide and tagged versions (v0–v8) to track the evolution of the agent.
  • Practical toolset: Includes bash execution, file reading/writing, pattern searching, and sub-agent spawning.
  • Safety-first: Implements tool risk classification and user permission confirmation for sensitive operations.
  • Persistent state: Supports session management, conversation compaction, and persistent memory storage.

Related

  • Project
    fitchmultz/pi-cursor-sdk

    pi‑cursor‑sdk is an open‑source bridge that runs the Cursor SDK agent locally and connects it to Pi’s terminal‑based AI assistant, letting users select Cursor models, use Cursor’s file/shell tools, and manage fast/plan modes directly from Pi.

  • Project
    drivelineresearch/autoresearch-claude-code

    autoresearch‑claude‑code is a Python framework that runs budget‑constrained auto‑research loops with Claude Code or Codex agents. You define a goal, a deterministic scorer, and files the agent may edit; the system creates a Git branch, measures each change, and automatically commits only statistically‑significant improvements while respecting run‑time and turn limits.

  • Project
    plmbr/notebook-intelligence

    Notebook Intelligence (NBI) is an open‑source AI assistant for JupyterLab that adds chat, inline code generation, auto‑completion, and programmable agents. It works with Copilot, OpenAI/LiteLLM endpoints, local Ollama models, or Anthropic Claude (via Claude Code). Features include a chat sidebar, inline suggestions, multi‑language awareness, a “Chatbook” mode for natural‑language cells, and various agent modes. Install with `pip install notebook-intelligence`, configure a provider in the NBI settings panel, and start chatting or generating code directly inside notebooks.

  • Project
    FalkorDB/code-graph

    CodeGraph is a tool that transforms codebases into interactive knowledge graphs stored in FalkorDB, enabling developers to visualize dependencies and query code structure using LLM-powered GraphRAG.