Every pattern is a real, runnable implementation — not a diagram or a blog post.

Install

pip install nanocoderagent

Pick your model — any OpenAI-compatible API works:

# Kimi K2.5
export OPENAI_API_KEY=your-key OPENAI_BASE_URL=https://api.moonshot.ai/v1
nanocoder -m kimi-k2.5

# Claude Opus 4.6 (via OpenRouter)
export OPENAI_API_KEY=your-key OPENAI_BASE_URL=https://openrouter.ai/api/v1
nanocoder -m anthropic/claude-opus-4-6

# OpenAI GPT-5
export OPENAI_API_KEY=sk-...
nanocoder -m gpt-5

# DeepSeek V3
export OPENAI_API_KEY=sk-... OPENAI_BASE_URL=https://api.deepseek.com
nanocoder -m deepseek-chat

# Qwen 3.5
export OPENAI_API_KEY=sk-... OPENAI_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
nanocoder -m qwen-max

# Ollama (local)
export OPENAI_API_KEY=ollama OPENAI_BASE_URL=http://localhost:11434/v1
nanocoder -m qwen3:32b

# One-shot mode
nanocoder -p "add error handling to parse_config()"

Architecture

The whole thing fits in your head:

nanocoder/
├── cli.py            REPL + commands               160 lines
├── agent.py          Agent loop + parallel tools    120 lines
├── llm.py            Streaming client + retry       150 lines
├── context.py        3-layer compression            145 lines
├── session.py        Save/resume                     65 lines
├── prompt.py         System prompt                   35 lines
├── config.py         Env config                      30 lines
└── tools/
    ├── bash.py       Shell + safety + cd tracking    95 lines
    ├── edit.py       Search-replace + diff            70 lines
    ├── read.py       File reading                     40 lines
    ├── write.py      File writing                     30 lines
    ├── glob_tool.py  File search                      35 lines
    ├── grep.py       Content search                   65 lines
    └── agent.py      Sub-agent spawning               50 lines

Use as a Library

from nanocoder import Agent, LLM

llm = LLM(model="kimi-k2.5", api_key="your-key", base_url="https://api.moonshot.ai/v1")
agent = Agent(llm=llm)
response = agent.chat("find all TODO comments in this project and list them")

Add Your Own Tools (~20 lines)

from nanocoder.tools.base import Tool

class HttpTool(Tool):
    name = "http"
    description = "Fetch a URL."
    parameters = {"type": "object", "properties": {"url": {"type": "string"}}, "required": ["url"]}

    def execute(self, url: str) -> str:
        import urllib.request
        return urllib.request.urlopen(url).read().decode()[:5000]

Commands

/model <name>    Switch model mid-conversation
/compact         Compress context (like Claude Code's /compact)
/tokens          Token usage
/save            Save session to disk
/sessions        List saved sessions
/reset           Clear history
quit             Exit

How It Compares

| | Claude Code | Claw-Code | Aider | NanoCoder | |---|---|---|---|---| | Code | 512K lines (closed) | 100K+ lines | 50K+ lines | 1,300 lines | | Models | Anthropic only | Multi | Multi | Any OpenAI-compatible | | Readable? | No | Hard | Medium | One afternoon | | Purpose | Use it | Use it | Use it | Understand it, build yours |

The Deep Dive

I wrote 7 articles breaking down Claude Code's architecture — the agent loop, tool system, context compression, streaming executor, multi-agent, and 44 hidden feature flags. If you want to understand why NanoCoder is designed this way, start there.

License

MIT. Fork it, learn from it, ship something better. A mention of this project is appreciated.

Built by Yufeng He · Agentic AI Researcher @ Moonshot AI (Kimi)

Claude Code Source Analysis — 170K+ reads, 6000 bookmarks on Zhihu