Nano-Collective/nanocoder: An open coding agent for your terminal, built by a community collective rather than a company. Bring your own model, keep your code on your machine, and owe nothing to anyone.

An open coding agent for your terminal, built by a community collective rather than a company. Bring your own model, keep your code on your machine, and owe nothing to anyone. - Nano-Collective/nan...

Summary

Nanocoder is a terminal-based coding agent that supports local models through Ollama or any OpenAI-compatible API, including OpenRouter, Requesty, Anthropic, and Google. It gives engineers control over model and provider selection while supporting interactive, non-interactive, review, and isolated worktree workflows.

What it is useful for

Nanocoder is useful for repository-level tasks that benefit from an AI coding agent without moving the workflow into a separate desktop application. Documented examples include analyzing a TypeScript file, refactoring a database module, and auditing an authentication module:

nanocoder --provider openrouter --model google/gemini-3.1-flash run "analyze src/app.ts"
nanocoder run --provider openrouter "refactor database module"
nanocoder --mode plan run "audit the auth module"

Its development modes include normal, auto-accept, yolo, plan, and architect, so an engineer can choose how much interaction or autonomy the session allows.

For GitHub-based work, nanocoder review main and nanocoder review 42 review a branch or pull request for bugs, security issues, and style violations. The worktree command provides a separate workspace for a pull request:

nanocoder worktree 42

This creates a sibling <repository>-pr-42 directory and a local nanocoder/pr-42 branch, leaving the original working tree untouched. It requires Git, an authenticated gh CLI, and an interactive terminal.

Nanocoder also documents skills, lifecycle hooks, checkpointing, task management, and MCP server configuration for engineers who need more structured project workflows. Terminal rendering is configurable: fullscreen mode is the default, while --no-alt-screen enables inline mode so completed messages remain in the terminal’s native scrollback.

How engineers can use it

Install the CLI with the documented npm command:

npm install -g @nanocollective/nanocoder
nanocoder

Then select a provider and model directly, for example with a local Ollama model:

nanocoder --provider ollama --model llama3.1

Provider, model, and mode flags can appear before or after the run command. Homebrew and Nix Flakes installations are also documented in the getting-started guide.

One limitation is that nanocoder review is currently a diff-only v1 feature. It requires a TTY and cannot be used with pipes or redirection; its output also cannot currently be captured to a file. Worktrees and branches created by nanocoder worktree remain after the session, so engineers must remove them with git worktree remove <path> and git branch -d nanocoder/pr-<number> when finished.