Making open source repositories work reliably for both the humans and the AI agents that now depend on them.

Open source is increasingly read and acted on by AI agents, not just people — but the signals a repository sends (its instructions, its examples, its dependency health, its release safety) were built for patient humans who improvise. Machine-operable open source is a research-and-tooling program that makes those signals checkable by machines, which helps the machines and the humans alike.

Each tool follows the same pattern: find a signal nobody checks, build the checker, and prove it matters with evidence.

• readme-ci — tests the code examples in your documentation, so "our examples always run" becomes a promise CI enforces instead of a hope. pip install readme-ci

More tools are on the way — agent-readable repository metadata, dependency lifecycle signals, and release crash-testing among them.

Evidence first. We measure a gap at ecosystem scale, ship a small open tool that closes it, and submit real fixes upstream — the intervention is the data. Everything is open source.

Learn more at machine-operable.org.