
Contributing to a Nano Collective project has no application process. Small, self-contained work (a bug fix, a docs typo, a dep bump, a refactor with no behaviour change) goes straight to a PR. Bigger ideas start with a Discord message or an issue. https://nanocollective.org



Every tool we ship is privacy-respecting, local-first, open for all. In that order, as a filter for what gets built and how. Not a tagline, an operating constraint. https://nanocollective.org



get-md ships filter flags for cleanup. --no-images, --no-links, --no-tables drop the noise. --no-frontmatter strips the YAML block. --no-extract keeps the full HTML structure. Compose what you want, skip what you don't. https://github.com/Nano-Collective/get-md



Old JSON files with no `_version` field are treated as version 0. `json-up` runs the full migration chain, so legacy data lands at the latest typed schema without a hand-rolled bootstrap step. https://github.com/Nano-Collective/json-up



Want to try a risky refactor without losing what works? Nanocoder's `/checkpoint create` and `/checkpoint load` save conversation history plus modified files, so a bad change is one command away from being undone. https://github.com/Nano-Collective/nanocoder



Nanotune benchmarks use five match modes: semantic, startsWith, contains, exact, and llm-judge. Pick the right one per test: code wants semantic, Q&A wants contains, labels want exact, open-ended wants llm-judge.

https://github.com/Nano-Collective/nanotune
