Nyxt provides introspective and help capabilities. All commands, classes, slots, variables, functions and bindings can be inspected for definition and documentation.

• tutorial (f1 t): Display Nyxt tutorial.
• describe-key (f1 k): Display binding of user-inputted keys.
• describe-bindings (f1 b): Show a buffer with the list of all known key bindings for the current buffer.
• describe-command (f1 c): Inspect a command and show it in a help buffer.
• describe-function (f1 f): Inspect a function and show it in a help buffer.
• describe-variable (f1 v): Inspect a variable and show it in a help buffer.
• describe-class (f1 C): Inspect a class and show it in a help buffer.
• describe-slot (f1 s): Inspect a slot and show it in a help buffer.
• describe-any (f1 a): Inspect anything and show it in a help buffer.

A good starting point is to study the documentation of the classes browser, window, buffer and prompt-buffer.

Nyxt is written in the Common Lisp programming language which offers a great perk: everything in the browser can be customized by the user, even while it's running!

To get started with Common Lisp, we recommend checking out our web page: Learn Lisp. It contains numerous pointers to other resources, including free books both for beginners and seasoned programmers.

Nyxt provides a mechanism for new users unfamiliar with Lisp to customize Nyxt. Start by invoking the commands describe-class (f1 C) or describe-slot (f1 s). You can press the button marked 'Configure' to change the value of a setting. The settings will be applied immediately and saved for future sessions. Please note that these settings will not alter existing object instances.

Settings created by Nyxt are stored in /home/doe/.config/nyxt/auto-config.3.lisp.

Any settings can be overridden manually by /home/doe/.config/nyxt/config.lisp.

The following section assumes knowledge of basic Common Lisp or a similar programming language.

The user needs to manually create the Nyxt configuration file, and the parent folders if necessary.

Example:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration web-buffer
  ((default-modes
    (pushnew 'nyxt/mode/no-script:no-script-mode %slot-value%))))

The above turns on the 'no-script-mode' (disables JavaScript) by default for every buffer.

The define-configuration macro can be used to customize the slots of classes like the browser, buffers, windows, etc. Refer to the class and slot documentation for the individual details.

To find out about all modes known to Nyxt, run describe-command (f1 c) and type 'mode'.

Slots store values that can be either accessed (get) or changed (set). Setting new values for slots allows many possibilities of customization. For instance, keyboard layouts vary across the world. The slot hints-alphabet has the default value of ABCDEFGHIJKLMNOPQRSTUVWXYZ. If the user has an American keyboard, they can do:

1. Execute command describe-slot (f1 s);
2. Type hints-alphabet;
3. Select hints-alphabet ( hint-mode class option);
4. Press the button Configure, and;
5. Insert the string "asfdghjkl"
.

This will make link-hinting more comfortable for this user. In addition, other similar approaches of customization can be applied to slots such as spell-check-language, which can be expanded to do the spelling-check of other languages besides English.

There are multiple buffer classes, such as document-buffer (for structured documents) and input-buffer (for buffers that can receive user input). A web-buffer class is used for web pages, prompt-buffer for, well, the prompt buffer. Some buffer classes may inherit from multiple other classes. For instance web-buffer and prompt-buffer both inherit from input-buffer.

You can configure one of the parent buffer classes slots and the new values will automatically cascade down as a new default for all child classes- unless this slot is specialized by these child classes. For instance if you configure the override-map slot in input-buffer, both panel-buffer and web-buffer classes will inherit from the new value.

Nyxt supports multiple bindings schemes such as CUA (the default), Emacs or vi. Changing scheme is as simple as setting the corresponding mode as default, e.g. emacs-mode. To make the change persistent across sessions, add the following to your configuration:

• vi bindings:

        
         Act on codeCopy
         Add To Auto-Config
         Try In Repl
        (define-configuration buffer
    ((default-modes
      (pushnew 'nyxt/mode/vi:vi-normal-mode %slot-value%))))

• Emacs bindings:

        
         Act on codeCopy
         Add To Auto-Config
         Try In Repl
        (define-configuration buffer
    ((default-modes
      (pushnew 'nyxt/mode/emacs:emacs-mode %slot-value%))))

You can create new scheme names with make-keyscheme. Also see the define-keyscheme-map macro.

To extend the bindings of a specific mode, you can configure the mode with define-configuration and extend its keyscheme-map with define-keyscheme-map. For example:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration base-mode
  "Note the :import part of the define-keyscheme-map.
It re-uses the other keymap (in this case, the one that was slot value before
the configuration) and merely adds/modifies it."
  ((keyscheme-map
    (define-keyscheme-map "my-base" (list :import %slot-value%)
                          keyscheme:vi-normal
                          (list "g b"
                                (lambda-command switch-buffer*
                                    nil
                                  (switch-buffer :current-is-last-p
                                                 t)))))))

The override-map is a keymap that has priority over all other keymaps. By default, it has few bindings like the one for execute-command (Ctrl+space). You can use it to set keys globally:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration input-buffer
  ((override-map
    (let ((map (make-keymap "override-map")))
      (define-key map "M-x" 'execute-command "C-space" 'nothing)))))

The nothing (unbound) command is useful to override bindings to do nothing. Note that it's possible to bind any command, including those of disabled modes that are not listed in execute-command (Ctrl+space). Binding to nothing (unbound) and binding to NIL means different things (see the documentation of define-key for details):

nothing (unbound)

Binds the key to a command that does nothing. Still discovers the key and recognizes it as pressed.

NIL

Un-binds the key, removing all the bindings that it had in a given mode/keyscheme-map. If you press the un-bound key, the bindings that used to be there will not be found anymore, and the key will be forwarded to the renderer.

Any other symbol/command

Replaces the command that was there before, with the new one. When the key is pressed, the new command will fire instead of the old one.

In addition, a more flexible approach is to create your own mode with your custom keybindings. When this mode is added first to the buffer mode list, its keybindings have priorities over the other modes. Note that this kind of global keymaps also have priority over regular character insertion, so you should probably not bind anything without modifiers in such a keymap.

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (defvar *my-keymap* (make-keymap "my-map"))

(define-key *my-keymap* "C-f" 'nyxt/mode/history:history-forwards
            "C-b" 'nyxt/mode/history:history-backwards)

(define-mode my-mode
    nil
  "Dummy mode for the custom key bindings in *my-keymap*."
  ((keyscheme-map
    (nkeymaps/core:make-keyscheme-map keyscheme:cua *my-keymap*
                                      keyscheme:emacs *my-keymap*
                                      keyscheme:vi-normal
                                      *my-keymap*))))

(define-configuration web-buffer
  "Enable this mode by default."
  ((default-modes (pushnew 'my-mode %slot-value%))))

Bindings are subject to various translations as per *translator*. By default if it fails to find a binding it tries again with inverted shifts. For instance if C-x C-F fails to match anything C-x C-f is tried.See the default value of *translator* to learn how to customize it or set it to nil to disable all forms of translation.

See the search-engines buffer slot documentation. Bookmarks can also be used as search engines, see the corresponding section.

Nyxt comes with default search engines for en.wikipedia.org, duckduckgo.com, search.atlas.engineer. The following example shows one way to add new search engines.

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (defvar *my-search-engines*
  (list
   '("google" "https://google.com/search?q=~a" "https://google.com")
   '("python3" "https://docs.python.org/3/search.html?q=~a"
     "https://docs.python.org/3")
   '("doi" "https://dx.doi.org/~a" "https://dx.doi.org/"))
  "List of search engines.")

(define-configuration context-buffer
  "Go through the search engines above and make-search-engine out of them."
  ((search-engines
    (append
     (mapcar (lambda (engine) (apply 'make-search-engine engine))
             *my-search-engines*)
     %slot-default%))))

Note that the last search engine is the default one. For example, in order to make python3 the default, the above code can be slightly modified as follows.

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (defvar *my-search-engines*
  (list
   '("google" "https://google.com/search?q=~a" "https://google.com")
   '("doi" "https://dx.doi.org/~a" "https://dx.doi.org/")
   '("python3" "https://docs.python.org/3/search.html?q=~a"
     "https://docs.python.org/3"))
  "List of search engines.")

(define-configuration context-buffer
  "Go through the search engines above and make-search-engine out of them."
  ((search-engines
    (append %slot-default%
            (mapcar
             (lambda (engine) (apply 'make-search-engine engine))
             *my-search-engines*)))))

If you don't want to use make-search-engine and want to try building the engines yourself, you can always make new search-engine and add it to search-engines list:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration context-buffer
  "Add a single search engine manually."
  ((search-engines
    (pushnew
     (make-instance 'search-engine :name "Reddit" :shortcut "r"
                    :search-url "https://reddit.com/search/?q=~a"
                    :fallback-url "https://reddit.com")
     %slot-value%))))

Nyxt history model is a tree whose nodes are URLs. It branches out through all the buffers. If you create a new buffer (via follow-hint-new-buffer (unbound) or make-buffer (unbound)), it becomes a new history branch originating from the branch of the previous buffer.

History can be navigated with the arrow keys in the status buffer, or with commands like history-backwards (unbound) and history-forwards (unbound) (which the arrows are bound to).

If the beyond-buffer-boundaries behavior sounds like too much to you, or you prefer the behavior of Nyxt 2, where the history was still a tree, but was not spilling across the buffers, then configure global-history-p to be NIL:

     
      Act on codeCopy
      Add To Auto-Config
      Try In Repl
     (define-configuration :context-buffer
  (global-history-p nil))

Nyxt supra-buffer history has benefits, though: it optimizes browsing patterns into more intuitive and productive structures. One particular pattern Nyxt history optimizes is hub-and-spoke search, where you keep returning to a certain hub to start your search/navigation from a familiar point. You can enable the optimization (merely going back in history to the hub page, instead of creating a new history node) for this strategy by configuring backtrack-to-hubs-p to T.

Another useful side to Nyxt tree-like history are braching-aware history commands, like history-forwards-query (unbound), allowing one to choose which branch of history they are going to visit, if there are several. If there's only one branch, then this command behaves much like regular history-forwards (unbound).

There are commands that allow to move across all the history before or after the current node:

• history-backwards-query (unbound): Query parent URL to navigate back to.
• history-forwards-all-query (unbound): Query URL to forward to, from all child branches.
• history-all-query (unbound): Query URL to go to, from the whole history.

If you need to know more: most of the optimizations and data structures are in history-tree library, while most of the Nyxt-specific interface is in nyxt/mode/history-tree.

See the list-downloads (Ctrl+Shift+Y) command and the download-path buffer slot documentation.

See the proxy-mode documentation.

This mode blocks access to websites related to specific hosts. To see all hosts being blocked, execute command describe-variable, choose variable NYXT/MODE/BLOCKER:*DEFAULT-HOSTLIST*, and read data on nyxt/mode/blocker:url-body slot. To customize host blocking, read the blocker-mode documentation.

You can configure which actions to take depending on the URL to be loaded. For instance, you can configure which Torrent program to start to load magnet links. See the url-dispatching-handler function documentation.

Auto-rules toggle modes when the URL satisfies the given conditions. URL-dispatchers can also be used for this, but it is simpler to use an auto-rule. Given that Nyxt's functionality is mode-based, the consequences are far reaching.

These can be used in the following ways:

• Manually, by calling:

save-non-default-modes-for-future-visits (unbound)

which saves "unusual" modes - non-default modes that were toggled exclusively for a given URL.

save-exact-modes-for-future-visits (unbound)

which saves the exact list of enabled modes for a given URL.

• Automatically, by setting prompt-on-mode-toggle-p to non-nil (refer to the configuration section for help).

All rules are stored at /home/doe/.local/share/nyxt/auto-rules.lisp, which is meant to be human-readable and human-writable. You can find instructions at the top of it. The gist is that rules are mere Lisp lists which start with a condition that checks the URL. When conditions are met, modes are toggled. Besides user-defined conditions, the following are often useful:

• match-domain
• match-host
• match-url
• match-regex
• match-scheme

By default, apply-all-matching-auto-rules-p is nil meaning that only the most specific rules are honored.

Auto-rules can also be defined for custom use-cases via define-auto-rule and un-defined with undefine-auto-rule.

Creating your own invocable commands is similar to creating a Common Lisp function, except the form is define-command instead of defun. If you want this command to be invocable outside of the context of a mode, use define-command-global.

Example:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-command-global my-bookmark-url
    nil
  "Query which URL to bookmark."
  (let ((url
         (prompt :prompt "Bookmark URL" :sources
                 'prompter:raw-source)))
    (nyxt/mode/bookmark:bookmark-add url)))

See the prompt-buffer class documentation for how to write custom prompt buffers.

You can also create your own context menu entries binding those to Lisp commands, using ffi-add-context-menu-command function. You can bind the bookmark-url like this:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (ffi-add-context-menu-command 'my-bookmark-url "Bookmark URL")

Currently, context menu commands don't have access to the renderer objects (and shouldn't hope to). Commands you bind to context menu actions should deduce most of the information from their surroundings, using JavaScript and Lisp functions Nyxt provides. For example, one can use the url-at-point to get thep URL currently under pointer.

With this, one can improve the bookmarking using url-at-point:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (ffi-add-context-menu-command
 (lambda ()
   (nyxt/mode/bookmark:bookmark-add (url-at-point (current-buffer))))
 "Bookmark Link")

If there's a scheme that Nyxt doesn't support, but you want it to, you can always define the handler for this scheme so that it's Nyxt-openable.

As a totally hypothetical example, you can define a nonsense scheme bleep to generate a page with random text:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-internal-scheme "bleep"
                        (lambda (url buffer)
                          (values
                           (spinneret:with-html-string
                             (:h1 "Bleep bloop?")
                             (:p
                              (loop repeat (parse-integer
                                            (quri.uri:uri-path
                                             (url url))
                                            :junk-allowed t)
                                    collect (:li
                                             (elt '("bleep" "bloop")
                                                  (random 2))))))
                           "text/html;charset=utf8"))
                        :local-p t)

What this piece of code does is

• Define a new scheme.
• Make a handler for it that takes the URL (as a string) and a buffer it's being opened in.
• Read the path (the part after the bleep:) of the URL and interpret it as a number.
• (Note that you need to wrap the URL into a url call so that it turns into a uri for the convenience of path (and other elements) fetching.)
• Generate a random list of "bleep" and "bloop".
• Return it as a text/html content.

The next time you run Nyxt and open bleep:20, you'll see a list of twenty bleeps and bloops.

Internal schemes can return any type of content (both strings and arrays of bytes are recognized), and they are capable of being CORS-enabled, protected, and are in general capable of whatever the renderer-provided schemes do.

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (define-internal-page-command-global random-number
    (&key (max 1000000))
    (buffer "*Random*")
  "Generates a random number on every reload."
  (spinneret:with-html-string
    (:h1 (princ-to-string (random max)))
    (:button.button :onclick
     (parenscript:ps
       (nyxt/parenscript:lisp-eval
        (:title "re-load/re-generate the random number")
        (reload-buffer buffer)))
     :title "Re-generate the random number again" "New number")))

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (define-internal-page not-a-command
    nil
    (:title "*Hello*" :page-mode 'base-mode)
  "Hello there!")

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (buffer-load-internal-page-focus 'not-a-command)

Hooks provide a powerful mechanism to tweak the behavior of various events that occur in the context of windows, buffers, modes, etc.

A hook holds a list of handlers. Handlers are named and typed functions. Each hook has a dedicated handler constructor.

Hooks can be 'run', that is, their handlers are run according to the combination slot of the hook. This combination is a function of the handlers. Depending on the combination, a hook can run the handlers either in parallel, or in order until one fails, or even compose them (pass the result of one as the input of the next). The handler types specify which input and output values are expected.

To add or delete a hook, you only need to know a couple of functions:

• handler a class to wrap hook handlers in.
• add-hook (also known as hooks:add-hook) allows you to add a handler to a hook,for it to be invoked when the hook fires.
• nhooks:on (also available as hooks:on) as a shorthand for the nhooks:add-hook.
• remove-hook (also available as hooks:remove-hook) that removes the handler from a certain hook.
• nhooks:once-on (also available as hooks:once-on) as a one-shot version of nhooks:on that removes the handler right after it's completed.

Many hooks are executed at different points in Nyxt, among others:

• Global hooks, such as after-init-hook or after-startup-hook.
• Window- or buffer-related hooks.
• window-make-hook for when a new window is created.
• window-delete-hook for when a window is deleted.
• window-set-buffer-hook for when the current-buffer changes in the window.
• buffer-load-hook for when there's a new page loading in the buffer.
• buffer-loaded-hook for when this page is mostly done loading (some scripts/image/styles may not be fully loaded yet, so you may need to wait a bit after it fires.)
• request-resource-hook for when a new request happens. Allows redirecting and blocking requests, and is a good place to do something conditioned on the links being loaded.
• prompt-buffer-ready-hook fires when the prompt buffer is ready for user input. You may need to call all-ready-p on the prompt to ensure all the sources it contains are ready too, and then you can safely set new inputs and select the necessary suggestions.
• Commands :before and :after methods.
• Try, for example, (defmethod set-url :after (&key (prefill-current-url-p t)) ...) to do something after the set-url finishes executing.
• Modes 'enable' and 'disable' methods and their :before, :after, and :around methods.
• Mode-specific hooks, like before-download-hook and after-download-hook for download.

For instance, if you want to force 'old.reddit.com' over 'www.reddit.com', you can set a hook like the following in your configuration file:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (defun old-reddit-handler (request-data)
  (let ((url (url request-data)))
    (setf (url request-data)
            (if (search "reddit.com" (quri.uri:uri-host url))
                (progn
                 (setf (quri.uri:uri-host url) "old.reddit.com")
                 (log:info "Switching to old Reddit: ~s"
                           (render-url url))
                 url)
                url)))
  request-data)

(define-configuration web-buffer
  ((request-resource-hook
    (sera:add-hook %slot-default% 'old-reddit-handler))))

(See url-dispatching-handler for a simpler way to achieve the same result.)

Or, if you want to set multiple handlers at once,

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration web-buffer
  ((request-resource-hook
    (reduce #'sera:add-hook '(old-reddit-handler auto-proxy-handler)
            :initial-value %slot-default%))))

Some hooks like the above example expect a return value, so it's important to make sure we return request-data here. See the documentation of the respective hooks for more details.

Nyxt provides a uniform configuration interface for all data files persisted to disk (bookmarks, cookies, etc.). To each file corresponds a nyxt-file object. An nyxt-profile is a customizable object that helps define general rules for data storage. Both nyxt-file and nyxt-profile compose, so it's possible to define general rules for all files (even for those not known in advance) while it's also possible to specialize some data given an nyxt-profile.

The profile can be set from command line and from the configuration file.You can list all known profiles (including the user-defined profiles) with the --list-profiles command-line option.

The nyxt-files can be passed a hint from the --with-file command line option, but each nyxt-file and profile rules are free to ignore it.

When a path ends with the .gpg extension, by default your GnuPG key is used to decrypt and encrypt the file transparently. Refer to the GnuPG documentation for how to set it up.

Note that the socket and the initialization nyxt-paths cannot be set in your configuration (the socket is used before the initialization file is loaded). Instead you can specify these paths from their respective command-line option. You can instantiate a unique, separate Nyxt instance when you provide a new socket path. This is particularly useful in combination with profiles, say to develop Nyxt or extensions.

Example to create a development profile that stores all data in /tmp/nyxt and stores bookmark in an encrypted file:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-class dev-profile (nyxt-profile)
              ((files:name :initform "nyxt-dev"))
              (:documentation "Development profile."))

(defmethod files:resolve ((profile dev-profile) (path nyxt-file))
  "Expand all data paths inside a temporary directory."
  (sera:path-join
   (files:expand (make-instance 'nyxt-temporary-directory))
   (uiop/pathname:relativize-pathname-directory (call-next-method))))

(defmethod files:resolve ((profile dev-profile) (file history-file))
  "Persist history to default location."
  (files:resolve (global-profile) file))

(define-configuration web-buffer
  "Make new profile the default."
  ((profile
    (make-instance
     (or (find-profile-class (getf *options* :profile))
         'dev-profile)))))

Then you can start a separate instance of Nyxt using this profile with nyxt --profile dev --socket /tmp/nyxt.socket.

Nyxt provides a uniform interface to some password managers including KeepassXC and Password Store. The supported installed password manager is automatically detected.See the password-interface buffer slot for customization.

You may use the define-configuration macro with any of the password interfaces to configure them. Please make sure to use the package prefixed class name/slot designators within the define-configuration.

• save-new-password (unbound): Query for name and new password to persist in the database.
• copy-password (unbound): Query password and save to clipboard.

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (defmethod initialize-instance :after
           ((interface password:keepassxc-interface)
            &key &allow-other-keys)
  "It's obviously not recommended to set master password here,
as your config is likely unencrypted and can reveal your password to someone
peeking at the screen."
  (setf (password:password-file interface)
          "/path/to/your/passwords.kdbx"
        (password:key-file interface) "/path/to/your/keyfile"
        (password:yubikey-slot interface) "1:1111"))

(define-configuration nyxt/mode/password:password-mode
  ((nyxt/mode/password:password-interface
    (make-instance 'password:keepassxc-interface))))

(define-configuration buffer
  ((default-modes
    (append (list 'nyxt/mode/password:password-mode) %slot-value%))))

Much of the visual style can be configured by the user. You can use the facilities provided by theme and browser theme slot. For example, to set a theme to a midnight-like one, you can add this snippet to your configuration file:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration browser
  ((theme
    (make-instance 'theme:theme :background-color "black"
                   :on-background-color "#808080" :accent-color
                   "#37a8e4" :on-accent-color "black" :primary-color
                   "gray" :on-primary-color "white" :secondary-color
                   "darkgray" :on-secondary-color "black")
    :doc
    "You can omit the colors you like in default theme, and they will stay as they were.")))

This, on the next restart of Nyxt, will repaint all the interface elements into a dark-ish theme.

As an alternative to the all-encompassing themes, you can alter the style of every individual class controlling Nyxt interface elements. All such classes have a style slot that you can configure with your own CSS like this:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration nyxt/mode/style:dark-mode
  ((style
    (theme:themed-css (theme *browser*)
      `(* :background-color ,theme:background "!important"
          :background-image none "!important" :color "red"
          "!important")
      `(a :background-color ,theme:background "!important"
        :background-image none "!important" :color "#AAAAAA"
        "!important")))))

This snippet alters the style of Nyxt dark mode to have a more theme-compliant colors, using the theme:themed-css macro (making all the theme colors you've configured earlier available as variables like theme:on-primary.)

You can evaluate code from the command line with --eval and --load. From a shell:

    
     Act on codeSpinneret::Copy
     Funcall
    $ nyxt --no-config --eval '+version+' 
  --load my-lib.lisp --eval '(format t "Hello ~a!~&" (my-lib:my-world))'

You can evaluate multiple --eval and --load in a row, they are executed in the order they appear.

You can also evaluate a Lisp file from the Nyxt interface with the load-file (Ctrl+O) command. For convenience, load-config-file (unbound) (re)loads your initialization file.

You can even make scripts. Here is an example foo.lisp:

    
     Act on codeSpinneret::Copy
     Funcall
    #!/bin/sh
#|
exec nyxt --script "$0"
|#

;; Your code follows:
(format t "~a~&" +version+)

--eval and --load can be commanded to operate over an existing instance instead of a separate instance that exits immediately.

The remote-execution-p of the remote instance must be non-nil:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration browser
  ((remote-execution-p t)))

To let know a private instance of Nyxt to load a foo.lisp script and run its foofunction:

    
     Act on codeSpinneret::Copy
     Funcall
    nyxt --profile nosave --remote --load foo.lisp --eval '(foo)' --quit

Note that --quitat the end of each Nyxt CLI call here. If you don't provide --quit when dealing with a remote instance, it will go into a REPL mode, allowing an immediate communication with an instance:

   nyxt --remote
(echo "~s" (+ 1 2)) ;; Shows '3' in the message area of remote Nyxt

User scripts are a conventional and lightweight way to run arbitrary JavaScript code on some set of pages/conditions. While not as powerful as either WebExtensions on Lisp-native extensions to Nyxt, those hook into the tenderer inner working and allow you to change the page and JavaScript objects associated to it.

As an example, you can remove navbars from all the pages you visit with this small configuration snippet (note that you'd need to have user-script-mode in your buffer default-modes ):

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (define-configuration web-buffer
  "Enable user-script-mode, if you didn't already."
  ((default-modes
    (pushnew 'nyxt/mode/user-script:user-script-mode %slot-value%))))

(define-configuration nyxt/mode/user-script:user-script-mode
  ((nyxt/mode/user-script:user-scripts
    (list
     (make-instance 'nyxt/mode/user-script:user-script :code
                    "// ==UserScript==
                              // @name          No navbars!
                              // @description	A simple script to remove navbars
                              // @run-at        document-end
                              // @include       http://*/*
                              // @include       https://*/*
                              // @noframes
                              // ==/UserScript==

                              var elem = document.querySelector(\"header\") || document.querySelector(\"nav\");
                              if (elem) {
                              elem.parentNode.removeChild(elem);
                              }"))
    :doc "Alternatively, save the code to some file and use
:base-path #p\"/path/to/our/file.user.js\".
Or fetch a remote script with
url (quri:uri \"https://example.com/script.user.js\")")))

Greasemonkey documentation lists all the possible properties that a user script might have. To Nyxt implementation, only those are meaningful:

@include and include

Sets the URL pattern to enable this script for. Follows the pattern scheme://host/path, where scheme is either a literal scheme or and asterisk (matching any scheme), and host and path are any valid characters plus asterisks (matching any set of characters) anywhere.

@match

Same as @include.

@exclude and exclude

Similar to @include, but rather disables the script for the matching pages.

@noframes and all-frames-p

When present, disables the script for all the frames but toplevel ones. When absent, injects the script everywhere. The Lisp-side all-frames-pworks in an opposite way.

@run-at and run-at

When to run a script. Allowed values: document-start, document-end, document-idle (in Nyxt implementation, same as document-end).

@require

Allows including arbitrary JS files hosted on the Internet or loaded from the same place as the script itself. Neat for including some JS libraries, like jQuery.

Similarly to Nyxt's scripting functionality, headless mode runs without a graphical user interface. Possible use-cases for this mode are web scraping, automations and web page analysis.

To enable headless mode, simply start Nyxt with the --headless CLI flag and provide a script file to serve as the configuration file:

    
     Act on codeSpinneret::Copy
     Funcall
    nyxt --headless --config /path/to/your/headless-config.lisp

Note that you pass it a configuration file—headless mode is only different from the regular Nyxt functions in that it has no GUI, and is all the same otherwise, contrary to all the seeming similarities to the --script flag usage.

The example below showcases frequent idioms that are found in the mode's configuration file:

    
     Act on codeSpinneret::Copy
     Funcall
    #!/bin/sh
#|
exec nyxt --headless --no-auto-config --profile nosave --config "$0"
|#

(define-configuration browser
  "Disable session restoration to speed up startup and get more reproducible
behavior."
  ((restore-session-on-startup-p nil)))

(define-configuration browser
  "Load the URL of Nyxt repository by default in all new buffers.
Alternatively, call buffer-load in after-startup-hook."
  ((default-new-buffer-url
    (quri.uri:uri "https://github.com/atlas-engineer/nyxt"))))

(hooks:on (after-startup-hook *browser*) (browser)
  ;; Once the page's done loading, do your thing.
  (hooks:once-on (buffer-loaded-hook (current-buffer)) (buffer)
    ;; It's sometimes necessary to sleep, as `buffer-loaded-hook' fires when the
    ;; page is loaded, which does not mean that all the resources and scripts
    ;; are done loading yet. Give it some time there.
    (sleep 0.5)
    ;; All the Nyxt reporting happens in headless mode, so you may want to log
    ;; it with `echo' and `echo-warning'.
    (echo "Nyxt GitHub repo open.")
    ;; Updating the `document-model' so that it includes the most relevant
    ;; information about the page.
    (nyxt:update-document-model)
    ;; Click the star button.
    (nyxt/dom:click-element
     (elt (clss:select "[aria-label=\"Star this repository\"]"
                       (document-model buffer))
                       0))
    (echo "Clicked the star.")
    ;; It's good tone to `nyxt:quit' after you're done, but if you use nyxt
    ;; --no-socket, you don't have to. Just be ready for some RAM eating :)
    (nyxt:quit)))

The contents of headless-config.lisp feature configuration forms that make Nyxt perform some actions to the opened pages and/or on certain hooks. Things you'd most probably want to put there are:

• Hook bindings, using the nhooks library and hooks provided by Nyxt.
• Operations on the page. Check the nyxt/dom library and the document-model method.
• The document-model method has a reasonably fresh copy of the page DOM (Document Object Model, reflecting the dynamic structure of the page). It is a Plump DOM, which means that all Plump (and CLSS) functions can be used on it.
• update-document-model is a function to force DOM re-parsing for the cases when you consider the current document-model too outdated.
• select is a CLSS function to find elements using CSS selectors (a terse notation for web page element description).
• clss:ordered-select is the same as select, except it guarantees that all the elements are returned in a depth-first traversal order.
• click-element to programmatically click a certain element (including the ones returned by select.)
• focus-select-element to focus an input field, for example.
• check-element to check a checkbox or a radio button.
• select-option-element to select an option from the <select> element options.

Additionally, headless mode gracefully interacts with other CLI toggles the Nyxt has:

• --headless itself! Notice that you can debug your script by omitting this CLI flag. When you're confident enough about it, put it back in. A good debugging tip, isn't it?
• --no-socket flag allows starting as many Nyxt instances as your machine can handle. Useful to parallelize computations.
• --profile nosave to not pollute your history and cache with the script-accessed pages.

Nyxt has a built-in REPL, available with repl (unbound) command.The REPL can be used to try out some code snippets for automation or quickly make some Lisp calculations. All the packages Nyxt depends on are available in REPL with convenient nicknames, and all the code is evaluated in nyxt-user package.

Once the REPL is open, there's only one input cell visible. This cell, always present at the bottom of the screen, adds new cells to the multi-pane interface of Nyxt REPL. You can type in (print "Hello, Nyxt!") and press C-return to evaluate the cell. A new cell will appear at the top of the buffer, with input area containing familiar code, with some v332 = "Hello, Nyxt!" variable assignment, and with a verbatim text outputted by your code:

   Hello, Nyxt!

This cell-based code evaluation is the basis of the Nyxt REPL. For more features, see REPL mode documentation.

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (define-class comment-cell (nyxt/mode/repl:cell)
              ((name "Commentary")) (:export-class-name-p t)
              (:export-accessor-names-p t))

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (defmethod nyxt/mode/repl:render-input ((cell comment-cell))
  "Render as a pre tag when ready, render as default input otherwise."
  (if (nyxt/mode/repl:ready-p cell)
      (spinneret:with-html-string
        (:pre (input cell)))
      (call-next-method)))

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (defmethod nyxt/mode/repl:render-results ((cell comment-cell))
  (declare (ignore cell))
  "")

      
       Act on codeCopy
       Add To Auto-Config
       Try In Repl
      (defmethod nyxt/mode/repl:evaluate ((cell comment-cell))
  (declare (ignore cell)))

While define-configuration is convenient, it is mostly restricted to class slot configuration. If you want to do anything else on class instantiation, you'll have to specialize the lower-level customize-instance generic function. Example:

    
     Act on codeCopy
     Add To Auto-Config
     Try In Repl
    (defmethod customize-instance ((buffer buffer) &key)
  (echo "Buffer ~a created." buffer))

All classes with metaclass user-class call customize-instance on instantiation, after initialize-instance :after. The primary method is reserved to the user, however the :after method is reserved to the Nyxt core to finalize the instance.

To install an extension, copy inside the *extensions-directory* (default to ~/.local/share/nyxt/extensions).

Extensions are regular Common Lisp systems.

A catalog of extensions is available in the document/EXTENSIONS.org file in the source repository.