Skip to content
LTI 1.3 Advantage Tool
Branch: master
Clone or download
Pull request Compare This branch is even with dmitry-viskov:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


LTI 1.3 Advantage Tool implementation in Python

PyPI Python Build Status MIT

This project is a Python implementation of the similar PHP tool. Library contains adapter for usage from Django Web Framework but there is no difficult to adapt it to from other frameworks: you should just re-implement OIDCLogin and MessageLaunch classes as it already done for Django.


First of all choose and configure test LTI 1.3 Platform. It may be:

The most simple way to check example is to use docker + docker-compose. Change the necessary configs in the examples/configs/game.json (here is instruction how to generate your own public + private keys):

    "<issuer>" : { // This will usually look something like ''
        "client_id" : "<client_id>", // This is the id received in the 'aud' during a launch
        "auth_login_url" : "<auth_login_url>", // The platform's OIDC login endpoint
        "auth_token_url" : "<auth_token_url>", // The platform's service authorization endpoint
        "key_set_url" : "<key_set_url>", // The platform's JWKS endpoint
        "key_set": null, // in case if platform's JWKS endpoint somehow unavailable you may paste JWKS here
        "private_key_file" : "<path_to_private_key>", // Relative path to the tool's private key
        "deployment_ids" : ["<deployment_id>"] // The deployment_id passed by the platform during launch

and execute:

$ docker-compose up --build

You may use virtualenv instead of docker:

$ virtualenv venv
$ source venv/bin/activate
$ cd examples
$ pip install -r requirements.txt
$ cd game
$ python runserver

Now there is game example tool you can launch into on the port 9001:

LTI Launch URL:

Configuration & Usage

Accessing Registration Data

To configure your own tool you may use built-in adapters:

from pylti1p3.tool_config import
tool_conf = ToolConfJsonFile('path/to/json')

from pylti1p3.tool_config import ToolConfDict
settings = {
    "<issuer_1>": { },
    "<issuer_2>": { }
private_key = '...'
tool_conf = ToolConfDict(settings)
tool_conf.set_private_key(iss, private_key)

or create your own implementation. The pylti1p3.tool_config.ToolConfAbstract interface must be fully implemented for this to work.

Open Id Connect Login Request

LTI 1.3 uses a modified version of the OpenId Connect third party initiate login flow. This means that to do an LTI 1.3 launch, you must first receive a login initialization request and return to the platform.

To handle this request, you must first create a new OIDCLogin (or DjangoOIDCLogin) object:

from pylti1p3.contrib.django import DjangoOIDCLogin

oidc_login = DjangoOIDCLogin(request, tool_conf)

Now you must configure your login request with a return url (this must be preconfigured and white-listed on the tool). If a redirect url is not given or the registration does not exist an pylti1p3.exception.OIDC_Exception will be thrown.

except OIDC_Exception:
    # display error page
    log.error('Error doing OIDC login')

With the redirect, we can now redirect the user back to the tool. There are three ways to do this:

This will add a HTTP 302 location header:


This will display some javascript to do the redirect instead of using a HTTP 302:

oidc_login.redirect(get_launch_url(request), js_redirect=True)

You can also get the url you need to redirect to, with all the necessary query parameters (if you would prefer to redirect in a custom way):

redirect_obj = oidc_login.get_redirect_object()
redirect_url = redirect_obj.get_redirect_url()

Redirect is done, we can move onto the launch.

LTI Message Launches

Now that we have done the OIDC log the platform will launch back to the tool. To handle this request, first we need to create a new MessageLaunch (or DjangoMessageLaunch) object.

message_launch = DjangoMessageLaunch(request, tool_conf)

Once we have the message launch, we can validate it. Validation is transparent - it's done once before you try to access the message body:

    launch_data = message_launch.get_launch_data()
except LtiException:
    log.error('Launch validation failed')

You may do it more explicitly:

    launch_data = message_launch.set_auto_validation(enable=False).validate()
except LtiException:
    log.error('Launch validation failed')

Now we know the launch is valid we can find out more information about the launch.

Check if we have a resource launch or a deep linking launch:

if message_launch.is_resource_launch():
    # Resource Launch!
elif message_launch.is_deep_link_launch():
    # Deep Linking Launch!
    # Unknown launch type

Check which services we have access to:

if message_launch.has_ags():
    # Has Assignments and Grades Service
if message_launch.has_nrps():
    # Has Names and Roles Service

Accessing Cached Launch Requests

It is likely that you will want to refer back to a launch later during subsequent requests. This is done using the launch id to identify a cached request. The launch id can be found using:

launch_id = message_launch.get_launch_id()

Once you have the launch id, you can link it to your session and pass it along as a query parameter.

Make sure you check the launch id against the user session to prevent someone from making actions on another person's launch.

Retrieving a launch using the launch id can be done using:

message_launch = DjangoMessageLaunch.from_cache(launch_id, request, tool_conf)

Once retrieved, you can call any of the methods on the launch object as normal, e.g.

if message_launch.has_ags():
    # Has Assignments and Grades Service

Deep Linking Responses

If you receive a deep linking launch, it is very likely that you are going to want to respond to the deep linking request with resources for the platform.

To create a deep link response you will need to get the deep link for the current launch:

deep_link = message_launch.get_deep_link()

Now we need to create pylti1p3.deep_link_resource.DeepLinkResource to return:

resource = DeepLinkResource()
    .set_custom_params({'my_param': my_param})\
    .set_title('My Resource')

Everything is set to return the resource to the platform. There are two methods of doing this.

The following method will output the html for an aut-posting form for you.

deep_link.output_response_form([resource1, resource2])

Alternatively you can just request the signed JWT that will need posting back to the platform by calling.

deep_link.get_response_jwt([resource1, resource2])

Names and Roles Service

Before using names and roles you should check that you have access to it:

if not message_launch.has_nrps():
    raise Exception("Don't have names and roles!")

Once we know we can access it, we can get an instance of the service from the launch.

nrps = message_launch.get_nrps()

From the service we can get list of all members by calling:

members = nrps.get_members()

Assignments and Grades Service

Before using assignments and grades you should check that you have access to it:

if not launch.has_ags():
    raise Exception("Don't have assignments and grades!")

Once we know we can access it, we can get an instance of the service from the launch:

ags = launch.get_ags()

To pass a grade back to the platform, you will need to create a pylti1p3.grade.Grade object and populate it with the necessary information:

gr = Grade()

To send the grade to the platform we can call:


This will put the grade into the default provided lineitem. If no default lineitem exists it will create one.

If you want to send multiple types of grade back, that can be done by specifying a pylti1p3.lineitem.LineItem:

line_item = LineItem()

ags.put_grade(gr, line_item);

If a lineitem with the same tag exists, that lineitem will be used, otherwise a new lineitem will be created.

You can’t perform that action at this time.