Skip to content

Built-in APIs

PyScript makes available convenience objects, functions and attributes.

In Python this is done via the builtin pyscript module:

Accessing the document object via the pyscript module
from pyscript import document

In HTML this is done via py-* and mpy-* attributes (depending on the interpreter you're using):

An example of a py-click handler
<button id="foo" py-click="handler_defined_in_python">Click me</button>

These APIs will work with both Pyodide and Micropython in exactly the same way.


Both Pyodide and MicroPython provide access to two further lower-level APIs:

  • Access to JavaScript's globalThis via importing the js module: import js (now js is a proxy for globalThis in which all native JavaScript based browser APIs are found).
  • Access to interpreter specific versions of utilities and the foreign function interface. Since these are different for each interpreter, and beyond the scope of PyScript's own documentation, please check each project's documentation (Pyodide / MicroPython) for details of these lower-level APIs.

PyScript can run in two contexts: the main browser thread, or on a web worker. The following three categories of API functionality explain features that are common for both main thread and worker, main thread only, and worker only. Most features work in both contexts in exactly the same manner, but please be aware that some are specific to either the main thread or a worker context.

Common features

These Python objects / functions are available in both the main thread and in code running on a web worker:


A Python dictionary representing the configuration for the interpreter.

Reading the current configuration.
from pyscript import config

# It's just a dict.


Changing the config dictionary at runtime has no effect on the actual configuration.

It's just a convenience to read the configuration at run time.


A utility function to retrieve the unique identifier of the element used to display content. If the element is not a <script> and it already has an id, that id will be returned.

The current_target utility
<!-- current_target(): explicit-id -->
<mpy-script id="explicit-id">
    from pyscript import display, current_target
    display(f"current_target(): {current_target()}")

<!-- current_target(): mpy-0 -->
    from pyscript import display, current_target
    display(f"current_target(): {current_target()}")

<!-- current_target(): mpy-1 -->
<!-- creates right after the <script>:
    <script-py id="mpy-1">
        <div>current_target(): mpy-1</div>
<script type="mpy">
    from pyscript import display, current_target
    display(f"current_target(): {current_target()}")


The return value of current_target() always references a visible element on the page, not at the current <script> that is executing the code.

To reference the <script> element executing the code, assign it an id:

<script type="mpy" id="unique-id">...</script>

Then use the standard document.getElementById(script_id) function to return a reference to it in your code.


A function used to display content. The function is intelligent enough to introspect the object[s] it is passed and work out how to correctly display the object[s] in the web page based on the following mime types:

  • text/plain to show the content as text
  • text/html to show the content as HTML
  • image/png to show the content as <img>
  • image/jpeg to show the content as <img>
  • image/svg+xml to show the content as <svg>
  • application/json to show the content as JSON
  • application/javascript to put the content in <script> (discouraged)

The display function takes a list of *values as its first argument, and has two optional named arguments:

  • target=None - the DOM element into which the content should be placed. If not specified, the target will use the current_script() returned id and populate the related dedicated node to show the content.
  • append=True - a flag to indicate if the output is going to be appended to the target.

There are some caveats:

  • When used in the main thread, the display function automatically uses the current <py-script> or <mpy-script> tag as the target into which the content will be displayed.
  • If the <script> tag has the target attribute, and is not a worker, the element on the page with that ID (or which matches that selector) will be used to display the content instead.
  • When used in a worker, the display function needs an explicit target="dom-id" argument to identify where the content will be displayed.
  • In both the main thread and worker, append=True is the default behaviour.
Various display examples
<!-- will produce
<py-script worker>
    from pyscript import display
    display("PyScript", append=False)

<!-- will produce
    <script type="py">...</script>
<script type="py">
    from pyscript import display
    display("PyScript", append=False)

<!-- will populate <h1>PyScript</h1> -->
<script type="py" target="my-h1">
    from pyscript import display
    display("PyScript", append=False)
<h1 id="my-h1"></h1>

<!-- will populate <h2>PyScript</h2> -->
<script type="py" worker>
    from pyscript import display
    display("PyScript", target="my-h2", append=False)
<h2 id="my-h2"></h2>


On both main and worker threads, this object is a proxy for the web page's document object. The document is a representation of the DOM and can be used to read or manipulate the content of the web page.


A common task is to fetch data from the web via HTTP requests. The pyscript.fetch function provides a uniform way to achieve this in both Pyodide and MicroPython. It is closely modelled on the Fetch API found in browsers with some important Pythonic differences.

The simple use case is to pass in a URL and await the response. Remember, in order to use await you must have the async attribute in the script tag that references your code. If this request is in a function, that function should also be defined as async.

A simple HTTP GET with pyscript.fetch
from pyscript import fetch

response = await fetch("")
if response.ok:
    data = await response.text()

The object returned from an await fetch call will have attributes that correspond to the JavaScript response object. This is useful for getting response codes, headers and other metadata before processing the response's data.

Alternatively, rather than using a double await (one to get the response, the other to grab the data), it's possible to chain the calls into a single await like this:

A simple HTTP GET as a single await
from pyscript import fetch

data = await fetch("").text()

The following awaitable methods are available to you to access the data returned from the server:

  • arrayBuffer() returns a Python memoryview of the response. This is equivalent to the arrayBuffer() method in the browser based fetch API.
  • blob() returns a JavaScript blob version of the response. This is equivalent to the blob() method in the browser based fetch API.
  • bytearray() returns a Python bytearray version of the response.
  • json() returns a Python datastructure representing a JSON serialised payload in the response.
  • text() returns a Python string version of the response.

The underlying browser fetch API has many request options that you should simply pass in as keyword arguments like this:

Supplying request options.
from pyscript import fetch

result = await fetch("", method="POST", body="HELLO").text()


You may encounter CORS errors (especially with reference to a missing Access-Control-Allow-Origin header.

This is a security feature of modern browsers where the site to which you are making a request will not process a request from a site hosted at another domain.

For example, if your PyScript app is hosted under and you make a request to (who don't allow requests from other domains) then you'll encounter this sort of CORS related error.

There is nothing PyScript can do about this problem (it's a feature, not a bug). However, you could use a pass-through proxy service to get around this limitation (i.e. the proxy service makes the call on your behalf).


The pyscript.ffi namespace contains foreign function interface (FFI) methods that work in both Pyodide and MicroPython.


A utility function explicitly for when a callback function is added via an event listener. It ensures the function still exists beyond the assignment of the function to an event. Should you not create_proxy around the callback function, it will be immediately garbage collected after being bound to the event.


There is some technical complexity to this situation, and we have attempted to create a mechanism where create_proxy is never needed.

Pyodide expects the created proxy to be explicitly destroyed when it's not needed / used anymore. However, the underlying proxy.destroy() method has not been implemented in MicroPython (yet).

To simplify this situation and automatically destroy proxies based on JavaScript memory management (garbage collection) heuristics, we have introduced an experimental flag:

experimental_create_proxy = "auto"

This flag ensures the proxy creation and destruction process is managed for you. When using this flag you should never need to explicitly call create_proxy.

The technical details of how this works are described here.


A utility function to convert Python references into their JavaScript equivalents. For example, a Python dictionary is converted into a JavaScript object literal (rather than a JavaScript Map), unless a dict_converter is explicitly specified and the runtime is Pyodide.

The technical details of how this works are described here.


It is possible to define JavaScript modules to use within your Python code.

Such named modules will always then be available under the pyscript.js_modules namespace.


Please see the documentation (linked above) about restrictions and gotchas when configuring how JavaScript modules are made available to PyScript.

The API wraps the browser's built-in IndexDB persistent storage in a synchronous Pythonic API.


The storage API is persistent per user tab, page, or domain, in the same way IndexedDB persists.

This API is not saving files in the interpreter's virtual file system nor onto the user's hard drive.

from pyscript import storage

# Each store must have a meaningful name.
store = await storage("my-storage-name")

# store is a dictionary and can now be used as such.

The returned dictionary automatically loads the current state of the referenced IndexDB. All changes are automatically queued in the background.

# This is a write operation.
store["key"] = value

# This is also a write operation (it changes the stored data).
del store["key"]

Should you wish to be certain changes have been synchronized to the underlying IndexDB, just await store.sync().

Common types of value can be stored via this API: bool, float, int, str and None. In addition, data structures like list, dict and tuple can be stored.


Because of the way the underlying data structure are stored in IndexDB, a Python tuple will always be returned as a Python list.

It is even possible to store arbitrary data via a bytearray or memoryview object. However, there is a limitation that such values must be stored as a single key/value pair, and not as part of a nested data structure.

Sometimes you may need to modify the behaviour of the dict like object returned by To do this, create a new class that inherits from pyscript.Storage, then pass in your class to as the storage_class argument:

from pyscript import window, storage, Storage

class MyStorage(Storage):

    def __setitem__(self, key, value):
        super().__setitem__(key, value)
        window.console.log(key, value)

store = await storage("my-data-store", storage_class=MyStorage)

# The store object is now an instance of MyStorage.


TODO: Use display(element) not element.display().

The classes and references in this namespace provide a Pythonic way to interact with the DOM. An explanation for how to idiomatically use this API can be found in the user guide


This object has two attributes and a single method:

  • head - a reference to a Python object representing the document's head.
  • body - a reference to a Python object representing the document's body.
  • find - a method that takes a single selector argument and returns a collection of Python objects representing the matching elements.

These are provided as a convenience so you have several simple and obvious options for accessing the content of the page (DOM).

All the Python objects returned by these attributes and method are instances of classes defined in the pyscript.web.elements namespace.


There are many classes in this namespace. Each is a one-to-one mapping of any HTML element name to a Python class representing the HTML element of that name. Each Python class ensures only valid properties and attributes can be assigned, according to web standards.

Usage of these classes is explained in the user guide.


The full list of supported element/class names is:

a, abbr, address, area, article, aside, audio
b, base, blockquote, body, br, button
canvas, caption, cite, code, col, colgroup
data, datalist, dd, del_, details, dialog, div, dl, dt
em, embed
fieldset, figcaption, figure, footer, form
h1, h2, h3, h4, h5, h6, head, header, hgroup, hr, html
i, iframe, img, input_, ins
label, legend, li, link
main, map_, mark, menu, meta, meter
object_, ol, optgroup, option, output
p, param, picture, pre, progress
s, script, section, select, small, source, span, strong, style, sub, summary, sup
table, tbody, td, template, textarea, tfoot, th, thead, time, title, tr, track
u, ul
var, video

These correspond to the standard HTML elements with the caveat that del_ and input_ have the trailing underscore (_) because they are also keywords in Python, and the grid is a custom class for a div with a grid style display property.

All these classes ultimately derive from the pyscript.web.elements.Element base class.

In addition to properties defined by the HTML standard for each type of HTML element (e.g. title, src or href), all elements have the following properties and methods (in alphabetical order):

  • append(child) - add the child element to the element's children.
  • children - a collection containing the element's child elements (that it contains).
  • classes - a set of CSS classes associated with the element.
  • clone(clone_id=None) - Make a clone of the element (and the underlying DOM object), and assign it the optional clone_id.
  • find(selector) - use a CSS selector to find matching child elements.
  • parent - the element's parent element (that contains it).
  • show_me - scroll the element into view.
  • style - a dictionary of CSS style properties associated with the element.
  • update(classes=None, style=None, **kwargs) - update the element with the specified classes (set), style (dict) and DOM properties (kwargs).
  • _dom_element - a reference to the proxy object that represents the underlying native HTML element.


All elements, by virtue of inheriting from the base Element class, may have the following properties:

accesskey, autofocus, autocapitalize,
className, contenteditable,
innerHTML, id,
part, popover,
slot, spellcheck,
tabindex, text, title, translate,

The classes set-like object has the following convenience functions:

  • add(*class_names) - add the class(es) to the element.
  • contains(class_name) - indicate if class_name is associated with the element.
  • remove(*class_names) - remove the class(es) from the element.
  • replace(old_class, new_class) - replace the old_class with new_class.
  • toggle(class_name) - add a class if it is absent, or remove a class if it is present.

Elements that require options (such as the datalist, optgroup and select elements), can have options passed in when they are created:

my_select = select_(option("apple", value=1), option("pear"))

Notice how options can be a tuple of two values (the name and associated value) or just the single name (whose associated value will default to the given name).

It's possible to access and manipulate the options of the resulting elements:

selected_option = my_select.options.selected
my_select.options.remove(0)  # Remove the first option (in position 0).

Finally, the collection of elements returned by find and children is iterable, indexable and sliceable:

for child in my_element.children[10:]:

Furthermore, four attributes related to all elements contained in the collection can be read (as a list) or set (applied to all contained elements):

  • classes - the list of classes associated with the elements.
  • innerHTML - the innerHTML of each element.
  • style - a dictionary like object for interacting with CSS style rules.
  • value - the value attribute associated with each element.


A Python decorator to indicate the decorated function should handle the specified events for selected elements.

The decorator takes two parameters:

  • The event_type should be the name of the browser event to handle as a string (e.g. "click").
  • The selector should be a string containing a valid selector to indicate the target elements in the DOM whose events of event_type are of interest.

The following example has a button with an id of my_button and a decorated function that handles click events dispatched by the button.

The HTML button
<button id="my_button">Click me!</button>
The decorated Python function to handle click events
from pyscript import when, display

@when("click", "#my_button")
def click_handler(event):
    Event handlers get an event object representing the activity that raised
    display("I've been clicked!")

This functionality is related to the py-* or mpy-* HTML attributes.


On the main thread, this object is exactly the same as import js which, in turn, is a proxy of JavaScript's globalThis object.

On a worker thread, this object is a proxy for the web page's global window context.


The reference for pyscript.window is always a reference to the main thread's global window context.

If you're running code in a worker this is not the worker's own global context. A worker's global context is always reachable via import js (the js object being a proxy for the worker's globalThis).


A class to wrap generic content and display it as un-escaped HTML on the page.

The HTML class
<script type="mpy">
    from pyscript import display, HTML

    # Escaped by default:
    display("<em>em</em>")  # &lt;em&gt;em&lt;/em&gt;

<script type="mpy">
    from pyscript import display, HTML

    # Un-escaped raw content inserted into the page:
    display(HTML("<em>em</em>"))  # <em>em</em>


This constant flag is True when the current code is running within a worker. It is False when the code is running within the main thread.


If a pyscript.fetch results in a call and response HTTP interaction with a web server, the pyscript.Websocket class provides a way to use websockets for two-way sending and receiving of data via a long term connection with a web server.

PyScript's implementation, available in both the main thread and a web worker, closely follows the browser's own WebSocket class.

This class accepts the following named arguments:

  • A url pointing at the ws or wss address. E.g.: WebSocket(url="ws://localhost:5037/")
  • Some protocols, an optional string or a list of strings as described here.

The WebSocket class also provides these convenient static constants:

  • WebSocket.CONNECTING (0) - the ws.readyState value when a web socket has just been created.
  • WebSocket.OPEN (1) - the ws.readyState value once the socket is open.
  • WebSocket.CLOSING (2) - the ws.readyState after ws.close() is explicitly invoked to stop the connection.
  • WebSocket.CLOSED (3) - the ws.readyState once closed.

A WebSocket instance has only 2 methods:

  • ws.send(data) - where data is either a string or a Python buffer, automatically converted into a JavaScript typed array. This sends data via the socket to the connected web server.
  • ws.close(code=0, reason="because") - which optionally accepts code and reason as named arguments to signal some specific status or cause for closing the web socket. Otherwise ws.close() works with the default standard values.

A WebSocket instance also has the fields that the JavaScript WebSocket instance will have:

  • binaryType - the type of binary data being received over the WebSocket connection.
  • bufferedAmount - a read-only property that returns the number of bytes of data that have been queued using calls to send() but not yet transmitted to the network.
  • extensions - a read-only property that returns the extensions selected by the server.
  • protocol - a read-only property that returns the name of the sub-protocol the server selected.
  • readyState - a read-only property that returns the current state of the WebSocket connection as one of the WebSocket static constants (CONNECTING, OPEN, etc...).
  • url - a read-only property that returns the absolute URL of the WebSocket instance.

A WebSocket instance can have the following listeners. Directly attach handler functions to them. Such functions will always receive a single event object.

  • onclose - fired when the WebSocket's connection is closed.
  • onerror - fired when the connection is closed due to an error.
  • onmessage - fired when data is received via the WebSocket. If the is a JavaScript typed array instead of a string, the reference it will point directly to a memoryview of the underlying bytearray data.
  • onopen - fired when the connection is opened.

The following code demonstrates a pyscript.WebSocket in action.

<script type="mpy" worker>
    from pyscript import WebSocket

    def onopen(event):

    def onmessage(event):

    def onclose(event):

    ws = WebSocket(url="ws://localhost:5037/")
    ws.onopen = onopen
    ws.onmessage = onmessage
    ws.onclose = onclose


It's also possible to pass in any handler functions as named arguments when you instantiate the pyscript.WebSocket class:

from pyscript import WebSocket

def onmessage(event):

ws = WebSocket(url="ws://", onmessage=onmessage)

Main-thread only features


A class used to instantiate a new worker from within Python.


Sometimes we disambiguate between interpreters through naming conventions (e.g. py or mpy).

However, this class is always PyWorker and the desired interpreter MUST be specified via a type option. Valid values for the type of interpreter are either micropython or pyodide.

The following fragments demonstrate how to evaluate the file on a new worker from within Python. - the file to run in the worker.
from pyscript import RUNNING_IN_WORKER, display, sync

display("Hello World", target="output", append=True)

# will log into devtools console
print(RUNNING_IN_WORKER)  # True
print("awake") - starts a new worker in Python.
from pyscript import PyWorker

# type MUST be either `micropython` or `pyodide`
PyWorker("", type="micropython")
The HTML context for the worker.
<script type="mpy" src="./">
<div id="output"></div>  <!-- The display target -->


The pyscript.workers reference allows Python code in the main thread to easily access named workers (and their exported functionality).

For example, the following Pyodide code may be running on a named worker (see the name attribute of the script tag):

<script type="py" worker name="py-version">
import sys

def version():
    return sys.version

# define what to export to main consumers
__export__ = ["version"]

While over on the main thread, this fragment of MicroPython will be able to access the worker's version function via the workers reference:

<script type="mpy" async>
from pyscript import workers

pyworker = await workers["py-version"]

# print the pyodide version
print(await pyworker.version())

Importantly, the workers reference will NOT provide a list of known workers, but will only await for a reference to a named worker (resolving when the worker is ready). This is because the timing of worker startup is not deterministic.

Should you wish to await for all workers on the page at load time, it's possible to loop over matching elements in the document like this:

<script type="mpy" async>
from pyscript import document, workers

for el in document.querySelectorAll("[type='py'][worker][name]"):
    await workers[el.getAttribute('name')]

# ... rest of the code

Worker only features


If a JavaScript module is only needed under certain circumstances, we provide an asynchronous way to import packages that were not originally referenced in your configuration.

A pyscript.js_import example.
<script type="py" async>
from pyscript import js_import, window

escaper, = await js_import("")


The js_import call returns an asynchronous tuple containing the JavaScript modules referenced as string arguments.



This is an experimental feature.

Feedback and bug reports are welcome!

If you have a lot of Python packages referenced in your configuration, startup performance may be degraded as these are downloaded.

If a Python package is only needed under certain circumstances, we provide an asynchronous way to import packages that were not originally referenced in your configuration.

A pyscript.py_import example.
<script type="py" async>
from pyscript import py_import

matplotlib, regex, = await py_import("matplotlib", "regex")

print(matplotlib, regex)

The py_import call returns an asynchronous tuple containing the Python modules provided by the packages referenced as string arguments.


A function used to pass serializable data from workers to the main thread.

Imagine you have this code on the main thread:

Python code on the main thread
from pyscript import PyWorker

def hello(name="world"):
    display(f"Hello, {name}")

worker = PyWorker("./")
worker.sync.hello = hello

In the code on the worker, you can pass data back to handler functions like this:

Pass data back to the main thread from a worker
from pyscript import sync


HTML attributes

As a convenience, and to ensure backwards compatibility, PyScript allows the use of inline event handlers via custom HTML attributes.


This classic pattern of coding (inline event handlers) is no longer considered good practice in web development circles.

We include this behaviour for historic reasons, but the folks at Mozilla have a good explanation of why this is currently considered bad practice.

These attributes, expressed as py-* or mpy-* attributes of an HTML element, reference the name of a Python function to run when the event is fired. You should replace the * with the actual name of an event (e.g. py-click or mpy-click). This is similar to how all event handlers on elements start with on in standard HTML (e.g. onclick). The rule of thumb is to simply replace on with py- or mpy- and then reference the name of a Python function.

A py-click event on an HTML button element.
<button py-click="handle_click" id="my_button">Click me!</button>
The related Python function.
from pyscript import window

def handle_click(event):
    Simply log the click event to the browser's console.

Under the hood, the pyscript.when decorator is used to enable this behaviour.


In earlier versions of PyScript, the value associated with the attribute was simply evaluated by the Python interpreter. This was unsafe: manipulation of the attribute's value could have resulted in the evaluation of arbitrary code.

This is why we changed to the current behaviour: just supply the name of the Python function to be evaluated, and PyScript will do this safely.