A markdown-it plugin that adds an id attribute to headings and optionally permalinks.

English | 中文 (v7.0.1)

This plugin adds an id attribute to headings, e.g. ## Foo becomes <h2 id="foo">Foo</h2>.

Optionally it can also include permalinks, e.g. <h2 id="foo"><a class="header-anchor" href="#foo">Foo</a></h2> and a bunch of other variants!

• Usage
• User-friendly URLs
• Manually setting the id attribute
• Compatible table of contents plugin
• Parsing headings from HTML blocks
• Browser example
• Permalinks
  • Header link
  • Link after header
  • Link inside header
  • ARIA hidden
  • Custom permalink
• Debugging
• Development

const md = require('markdown-it')()
  .use(require('markdown-it-anchor'), opts)

See a demo as JSFiddle.

The opts object can contain:

All headers greater than the minimum level will have an id attribute with a slug of their content. For example, you can set level to 2 to add anchors to all headers but h1. You can also pass an array of header levels to apply the anchor, like [2, 3] to have an anchor on only level 2 and 3 headers.

If a permalink renderer is given, it will be called for each matching header to add a permalink. See permalinks below.

If a slugify function is given, you can decide how to transform a heading text to a URL slug. See user-friendly URLs.

The callback option is a function that will be called at the end of rendering with the token and an info object. The info object has title and slug properties with the token content and the slug used for the identifier.

We set by default tabindex="-1" on headers. This marks the headers as focusable elements that are not reachable by keyboard navigation. The effect is that screen readers will read the title content when it's being jumped to. Outside of screen readers, the experience is the same as not setting that attribute. You can override this behavior with the tabIndex option. Set it to false to remove the attribute altogether, otherwise the value will be used as attribute value.

Finally, you can customize how the title text is extracted from the markdown-it tokens (to later generate the slug). See user-friendly URLs.

Starting from v5.0.0, markdown-it-anchor dropped the string package to retain our core value of being an impartial and secure library. Nevertheless, users looking for backward compatibility may want the old slugify function:

npm install string

const string = require('string')
const slugify = s => string(s).slugify().toString()

const md = require('markdown-it')()
  .use(require('markdown-it-anchor'), { slugify })

Another popular library for this is @sindresorhus/slugify, which have better Unicode support and other cool features:

npm install @sindresorhus/slugify

const slugify = require('@sindresorhus/slugify')

const md = require('markdown-it')()
  .use(require('markdown-it-anchor'), { slugify: s => slugify(s) })

Additionally, if you want to further customize the title that gets passed to the slugify function, you can do so by customizing the getTokensText function, that gets the plain text from a list of markdown-it inline tokens:

function getTokensText (tokens) {
  return tokens
    .filter(token => !['html_inline', 'image'].includes(token.type))
    .map(t => t.content)
    .join('')
}

const md = require('markdown-it')()
  .use(require('markdown-it-anchor'), { getTokensText })

By default we include only text and code_inline tokens, which appeared to be a sensible approach for the vast majority of use cases.

An alternative approach is to include every token's content except for html_inline and image tokens, which yields the exact same results as the previous approach with a stock markdown-it, but would also include custom tokens added by any of your markdown-it plugins, which might or might not be desirable for you. Now you have the option!

If you need access to the markdown-it state from the slugify function, e.g. to access state.env, you can use slugifyWithState instead.

const md = require('markdown-it')()
  .use(require('markdown-it-anchor'), {
    slugifyWithState: (title, state) => `${state.env.id}-${slugify(title)}`
  })

You might want to explicitly set the id attribute of your headings from the Markdown document, for example to keep them consistent across translations.

markdown-it-anchor is designed to reuse any existing id, making markdown-it-attrs a perfect fit for this use case. Make sure to load it before markdown-it-anchor!

Then you can do something like this:

# Your title {#your-custom-id}

The anchor link will reuse the id that you explicitly defined.

Looking for an automatic table of contents (TOC) generator? Take a look at markdown-it-toc-done-right it's made from the ground to be a great companion of this plugin.

markdown-it-anchor doesn't parse HTML blocks, so headings defined in HTML blocks will be ignored. If you need to add anchors to both HTML headings and Markdown headings, the easiest way would be to do it on the final HTML rather than during the Markdown parsing phase:

const { parse } = require('node-html-parser')

const root = parse(html)

for (const h of root.querySelectorAll('h1, h2, h3, h4, h5, h6')) {
  const slug = h.getAttribute('id') || slugify(h.textContent)
  h.setAttribute('id', slug)
  h.innerHTML = `<a href="#${slug}">${h.innerHTML}</a>`
}

console.log(root.toString())

Or with a (not accessible) GitHub-style anchor, replace the h.innerHTML part with:

h.insertAdjacentHTML('afterbegin', `<a class="anchor" aria-hidden="true" href="#${slug}">🔗</a> `)

While this still needs extra work like handling duplicated slugs and IDs, this should give you a solid base.

That said if you really want to use markdown-it-anchor for this even though it's not designed to, you can do like npm does with their marky-markdown parser, and transform the html_block tokens into a sequence of heading_open, inline, and heading_close tokens that can be handled by markdown-it-anchor:

const md = require('markdown-it')()
  .use(require('@npmcorp/marky-markdown/lib/plugin/html-heading'))
  .use(require('markdown-it-anchor'), opts)

While they use regexes to parse the HTML and it won't gracefully handle any arbitrary HTML, it should work okay for the happy path, which might be good enough for you.

You might also want to check this implementation which uses Cheerio for a more solid parsing, including support for HTML attributes.

The only edge cases I see it failing with are multiple headings defined in the same HTML block with arbitrary content between them, or headings where the opening and closing tag are defined in separate html_block tokens, both which should very rarely happen.

If you need a bulletproof implementation, I would recommend the first HTML parser approach I documented instead.

See example.html.

Version 8.0.0 completely reworked the way permalinks work in order to offer more accessible options out of the box. You can also make your own permalink.

Instead of a single default way of rendering permalinks (which used to have a poor UX on screen readers), we now have multiple styles of permalinks for you to chose from.

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink[styleOfPermalink](permalinkOpts)
})

Here, styleOfPermalink is one of the available styles documented below, and permalinkOpts is an options object.

All renderers share a common set of options:

For the symbol, you may want to use the link symbol, or a symbol from your favorite web font.

This style wraps the header itself in an anchor link. It doesn't use the symbol option as there's no symbol needed in the markup (though you could add it with CSS using ::before if you like).

It's so simple it doesn't have any behaviour to custom, and it's also accessible out of the box without any further configuration, hence it doesn't have other options than the common ones described above.

You can find this style on the MDN as well as HTTP Archive and their Web Almanac, which to me is a good sign that this is a thoughtful way of implementing permalinks. This is also the style that I chose for my own blog.

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.headerLink()
})

<h2 id="title"><a class="header-anchor" href="#title">Title</a></h2>

The main caveat of this approach is that you can't include links inside headers. If you do, consider the other styles.

Also note that this pattern breaks reader mode in Safari, an issue you can also notice on the referenced websites above. This was already reported to Apple but their bug tracker is not public. In the meantime, a fix mentioned in the article above is to insert a span inside the link. You can use the safariReaderFix option to enable it.

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.headerLink({ safariReaderFix: true })
})

<h2 id="title"><a class="header-anchor" href="#title"><span>Title</span></a></h2>

If you want to customize further the screen reader experience of your permalinks, this style gives you much more freedom than the header link.

It works by leaving the header itself alone, and adding the permalink after it, giving you different methods of customizing the assistive text. It makes the permalink symbol aria-hidden to not pollute the experience, and leverages a visuallyHiddenClass to hide the assistive text from the visual experience.

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.linkAfterHeader({
    style: 'visually-hidden',
    assistiveText: title => `Permalink to “${title}”`,
    visuallyHiddenClass: 'visually-hidden',
    wrapper: ['<div class="wrapper">', '</div>']
  })
})

<div class="wrapper">
  <h2 id="title">Title</h2>
  <a class="header-anchor" href="#title">
    <span class="visually-hidden">Permalink to “Title”</span>
    <span aria-hidden="true">#</span>
  </a>
</div>

By using a visually hidden element for the assistive text, we make sure that the assistive text can be picked up by translation services, as most of the popular translation services (including Google Translate) currently ignore aria-label.

If you prefer an alternative method for the assistive text, see other styles:

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.linkAfterHeader({
    style: 'aria-label'
    assistiveText: title => `Permalink to “${title}”`
  })
})

<h2 id="title">Title</h2>
<a class="header-anchor" href="#title" aria-label="Permalink to “Title”">#</a>

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.linkAfterHeader({
    style: 'aria-describedby' // Or `aria-labelledby`
  })
})

<h2 id="title">Title</h2>
<a class="header-anchor" href="#title" aria-describedby="title">#</a>

This is the equivalent of the default permalink in previous versions. The reason it's not the first one in the list is because this method has accessibility issues.

If you use a symbol like just # without adding any markup around, screen readers will read it as part of every heading (in the case of #, it could be read "pound", "number" or "number sign") meaning that if you title is "my beautiful title", it will read "number sign my beautiful title" for example. For other common symbols, 🔗 is usually read as "link symbol" and ¶ as "pilcrow".

Additionally, screen readers users commonly request the list of all links in the page, so they'll be flooded with "number sign, number sign, number sign" for each of your headings.

I would highly recommend using one of the markups above which have a better experience, but if you really want to use this markup, make sure to pass accessible HTML as symbol to make things usable, like in the example below, but even that has some flaws.

With that said, this permalink allows the following options:

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.linkInsideHeader({
    symbol: `
      <span class="visually-hidden">Jump to heading</span>
      <span aria-hidden="true">#</span>
    `,
    placement: 'before'
  })
})

<h2 id="title">
  <a class="header-anchor" href="#title">
    <span class="visually-hidden">Jump to heading</span>
    <span aria-hidden="true">#</span>
  </a>
  Title
</h2>

While this example allows more accessible anchors with the same markup as previous versions of markdown-it-anchor, it's still not ideal. The assistive text for permalinks will be read as part of the heading when listing all the titles of the page, e.g. "jump to heading title 1, jump to heading title 2" and so on. Also that assistive text is not very useful when listing the links in the page (which will read "jump to heading, jump to heading, jump to heading" for each of your permalinks).

This is just an alias for linkInsideHeader with ariaHidden: true by default, to mimic GitHub's way of rendering permalinks.

Setting aria-hidden="true" makes the permalink explicitly inaccessible instead of having the permalink and its symbol being read by screen readers as part of every single headings (which was a pretty terrible experience).

const anchor = require('markdown-it-anchor')
const md = require('markdown-it')()

md.use(anchor, {
  permalink: anchor.permalink.ariaHidden({
    placement: 'before'
  })
})

<h2 id="title"><a class="header-anchor" href="#title" aria-hidden="true">#</a> Title</h2>

While no experience might be arguably better than a bad experience, I would instead recommend using one of the above renderers to provide an accessible experience. My favorite one is the header link, which is also the simplest one.

If none of those options suit you, you can always make your own renderer! Take inspiration from the code behind all permalinks.

The signature of the function you pass in the permalink option is the following:

function renderPermalink (slug, opts, state, idx) {}

Where opts are the markdown-it-anchor options, state is a markdown-it StateCore instance, and idx is the index of the heading_open token in the state.tokens array. That array contains Token objects.

To make sense of the "token stream" and the way token objects are organized, you will probably want to read the markdown-it design principles page.

This function can freely modify the token stream (state.tokens), usually around the given idx, to construct the anchor.

Because of the way the token stream works, a heading_open token is usually followed by a inline token that contains the actual text (and inline markup) of the heading, and finally a heading_close token. This is why you'll see most built-in permalink renderers touch state.tokens[idx + 1], because they update the contents of the inline token that follows a heading_open.

If you want to debug this library more easily, we support source maps.

Use the source-map-support module to enable it with Node.js.

node -r source-map-support/register your-script.js

# Build the library in the `dist/` directory.
npm run build

# Watch file changes to update `dist/`.
npm run dev

# Run tests, will use the build version so make sure to build after
# making changes.
npm test