Your First Plugin

Plugins are your tool for customizing the Canvas experience. By using the modules of the Canvas SDK, you can react to events emitted from the EHR, request additional data if needed, and respond with effects that alter workflows and add or change data in Canvas. You can also use utils to do things like call out to web services with our provided HTTP client.

Video #

The video below showcases a Canvas engineer working through this guide step-by-step.

1. Install the Canvas CLI #

To install the Canvas CLI, simply pip install canvas. Python 3.11 or 3.12 is required. You can find additional detail on the features of the Canvas CLI here.

2. Configure the Canvas CLI for your instances #

The Canvas CLI uses OAuth credentials to connect to your Canvas instance. If you’ve used our FHIR API, you’ll be very familiar with the process for registering credentials. Register a separate OAuth application, choosing confidential for the Client type, and client-credentials for the Authorization grant type. Redirect URIs can be left blank, and the Algorithm should be No OIDC support. Note the client_id and client_secret for the next step.

Create a file at the path ~/.canvas/credentials.ini. Here is what its contents should look like:



Each section represents credentials for a different Canvas instance. Replace the section headers with your Canvas subdomains. The example configuration provided would be valid for instances with URLs and

You can optionally set the is_default flag for the instance you wish to be implied when using the CLI. If no section is set as default, the first one will be considered default.

3. Initialize a new plugin #

The Canvas CLI gives you a great head start when creating a plugin. Simply run canvas init, and answer the prompt to name your plugin.

$ canvas init
  [1/1] project_name (My Cool Plugin): Paperwork Eviscerator
Project created in /Users/andrew/src/canvas-plugins/paperwork_eviscerator

This output shows the location of our freshly generated plugin.

4. Navigate the structure of a plugin #

Let’s take a look at what was generated for us.

$ tree paperwork_eviscerator/
`-- protocols

2 directories, 4 files


The CANVAS_MANIFEST.json is particularly important. It is used during the installation of the plugin.

    "sdk_version": "0.1.4",
    "plugin_version": "0.0.1",
    "name": "paperwork_eviscerator",
    "description": "Edit the description in CANVAS_MANIFEST.json",
    "components": {
        "protocols": [
                "class": "paperwork_eviscerator.protocols.my_protocol:Protocol",
                "description": "A protocol that does xyz...",
                "data_access": {
                    "event": "",
                    "read": [],
                    "write": []
        "commands": [],
        "content": [],
        "effects": [],
        "views": []
    "secrets": ["my_secret_code"],
    "tags": {},
    "references": [],
    "license": "",
    "diagram": false,
    "readme": "./"

The name, plugin version, and description are all surfaced in your Canvas instance when viewing installed plugins.

Only protocols declared here are invoked by the plugin runner. If they are not declared, they will be ignored.

Secrets can be declared (though not defined) here. Any secrets declared here will be initialized on plugin install, and can be set in the plugin listing in the Settings section of your Canvas instance. #

Share details about the purpose of your plugins and how it works in this README file.

protocols/ #

This file contains the protocol class declared in the manifest file. We’ve included some sample content and copious comments for inspiration.

from import EventType
from canvas_sdk.protocols import BaseProtocol
from logger import log

# Inherit from BaseProtocol to properly get registered for events
class Protocol(BaseProtocol):
    You should put a helpful description of this protocol's behavior here.

    # Name the event type you wish to run in response to

    NARRATIVE_STRING = "I was inserted from my plugin's protocol."

    def compute(self):
        This method gets called when an event of the type RESPONDS_TO is fired.
        # This class is initialized with several pieces of information you can
        # access.
        # `self.event` is the event object that caused this method to be
        # called.
        # `` is an identifier for the object that is the subject of
        # the event. In this case, it would be the identifier of the assess
        # command. If this was a patient create event, it would be the
        # identifier of the patient. If this was a task update event, it would
        # be the identifier of the task. Etc, etc.
        # `self.context` is a python dictionary of additional data that was
        # given with the event. The information given here depends on the
        # event type.
        # `self.secrets` is a python dictionary of the secrets you defined in
        # your CANVAS_MANIFEST.json and set values for in the uploaded
        # plugin's configuration page: <emr_base_url>/admin/plugin_io/plugin/<plugin_id>/change/
        # Example: self.secrets['WEBHOOK_URL']

        # You can log things and see them using the Canvas CLI's log streaming
        # function.

        # Craft a payload to be returned with the effect(s).
        payload = {
            "note": {"uuid": self.context["note"]["uuid"]},
            "data": {"narrative": self.NARRATIVE_STRING},

        # Return zero, one, or many effects.
        # Example:
        # return [Effect(type=EffectType.LOG, payload=json.dumps(payload))]
        return []

5. Listen for an Event #

Set the RESPONDS_TO value to the Event Type you’re interested in.

6. Return an Effect #

Form an Effect to return to your Canvas instance.

7. Deploy and use your plugin #

When your plugin is just the way you’d like it, deploying is simple. Simply run canvas install <path/to/plugin_root> and your plugin will be packaged, uploaded, installed, and enabled. As you make changes to your plugin, run the same command to update the code of the installed plugin.