Paste a range of commit messages to see which conform to Conventional Commits, what each one is missing, and what version bump the whole set implies.
Waiting for commit messages
One per line, or separate multi-line messages with a blank line. `git log --format=%B` gives you the right shape.
The findings appear here
Each message with what conforms, what does not, and what the set implies for the next version.
The implied release is what the convention says these messages mean. It is not what will happen — that depends on your release tooling actually reading conventional commits, and on the version you are starting from. Below 1.0.0 the specification is silent, and most tools treat a breaking change as a minor bump rather than a major one.
Commit Message Validator runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
Checking one commit message is something your editor can do. The question worth asking before a release is what a whole range of them adds up to — and the answer is frequently a major bump that nobody noticed, because one `fix!` is in there among forty ordinary commits.
`git log --format=%B v1.1.0..HEAD` gives you exactly the shape this expects. Paste it and every message is parsed, validated and summarised, with the implied release computed from the set rather than from any one entry.
Errors, warnings and notes are kept apart because they are not the same kind of thing. An error is a message the specification rejects outright. A warning is something that will cause a problem — a header past the limit, a capitalised description, a type nothing else uses. A note is a preference. A linter that reports all three at the same volume is a linter that gets switched off.
Messages that do not parse as conventional are counted and listed rather than ignored, because they are the ones that silently do not reach the changelog. A release computed from a history that is only half conventional is a guess, and the number of unparsed messages is how you tell whether it is a good one.
Nothing is uploaded. The messages are parsed in the page, which matters when a commit body mentions an internal system or a customer.
`git log --format=%B` output. Messages are separated by blank lines, which is how multi-line bodies stay attached to their headers. A plain list of one-line messages works too.
Usually because of commits that did not parse — they contribute nothing here and nothing in your tool either, so a missing feature is a missing minor bump. Below 1.0.0 is the other common reason: most tools treat a breaking change there as a minor rather than a major.
It is a convention rather than a rule. It comes from leaving room for git's own indentation in `git log` output within an 80-column terminal. Nothing breaks past it; GitHub truncates the subject in some views around 72, which is the practical reason it persists.
No. The specification makes it optional, and in a single-package repository it usually adds noise. In a monorepo it is how a changelog gets grouped usefully, which is why the requirement is a switch rather than a default.
No. They are parsed in your browser and never leave the page.