A coding agent that answers to no company

What it does

Nanocoder is a terminal-based coding agent that hooks up to whatever LLM you already trust—local models through Ollama, or remote endpoints like OpenRouter, Anthropic, and Google. It reads your codebase, executes commands, and edits files, aiming for the same agentic loop as Claude Code or Gemini CLI while keeping the work on your machine if you choose.

The interesting bit

The unusual part isn’t the feature set; it’s the governance. The project is built by the Nano Collective, a not-for-profit community that explicitly bans paid tiers, closed-source gatekeeping, and quiet telemetry. They even publish an Economics Charter to spell out how contributors get paid for bounties without turning the tool into a product.

Key highlights

• Bring-your-own-model support: Ollama for local inference, or any OpenAI-compatible API.
• Runs in multiple development modes (normal, auto-accept, plan, and yolo) alongside a per-project daemon and checkpointing.
• Explicitly local-first and privacy-respecting: no telemetry shipping prompts to hidden endpoints.
• No paid tiers or closed-source feature gating; the collective even publishes an Economics Charter for contributor bounties.

Verdict

Worth a look if you want a Claude Code alternative that stays local-first and won’t pivot into a SaaS next quarter. Skip it if you need the polish and dedicated support that comes with a funded company behind the tool.