Markdown Link Check

About this tool

Markdown Link Check scans a pasted Markdown document and reports link problems that are fully decidable offline. It does not fetch URLs, read your filesystem or upload the document anywhere; the same Rust engine runs in the CLI, chat block and browser page.

Use it before publishing a README, changelog or docs page to catch the mistakes that normal spell checkers miss:

• malformed inline links such as [text] (url), (text)[url], unclosed ( and empty () targets;
• undefined, duplicate or unused reference-style definitions;
• in-document #anchors that do not match any heading, custom {#id} or HTML id="..." anchor;
• empty link text, missing image alt text, unencoded spaces in destinations and malformed mailto: links;
• optional http:// hygiene warnings.

Fenced code blocks and inline code spans are ignored, so examples in snippets do not create noise. Line and column numbers are 1-based.

Worked example

Input:

# Install

See [setup](#setup), [site](https://example.com), and [missing][ref].

[ref]: https://example.com/one
[ref]: https://example.com/two

Output:

Issues
  3:5  error  ML007  broken anchor '#setup' — no heading in this document produces that id
  6:1  error  ML005  duplicate reference definition [ref] — the first definition wins, this one is ignored

2 error(s), 0 warning(s) in 3 link(s) checked.

Controls

• Link kind filters the report to anchors, external URLs, relative paths, mail links, images, empty targets or reference links.
• Report format chooses plain text, a Markdown table for issue comments, or JSON for scripts.
• Also list passing links adds every checked link to the report, not just the findings.
• Check in-document anchors can be disabled when your renderer uses custom slug rules.
• Warn about http:// links is off by default because some docs deliberately cite legacy URLs.

Limits and edge cases

• Maximum input size is 1 MB.
• External URLs are not fetched; this tool reports structural and hygiene issues, not HTTP status codes.
• Relative links are classified but not checked on disk, because the browser/wasm surface has no project filesystem.
• Heading anchors follow GitHub-style slugs, including duplicate headings as -1, -2; custom {#id} and raw HTML id="..." / name="..." anchors are also recognized.
• Reference labels are case-insensitive and whitespace-normalized.

FAQ