This guide covers setting up an MCP (Model Context Protocol) server in Workato, connecting it to a Totara site, and using it from an AI tool such as Microsoft Copilot or ChatGPT.
Note: This is a technical implementation guide, intended for whoever configures the integration (for example, a site administrator, developer, or implementation partner) rather than general Totara users. It assumes familiarity with Workato.
The example in this guide builds a single tool that retrieves compliance data from a Totara report for an AI agent to use. Microsoft Copilot and ChatGPT are covered as example AI platforms; MCP is a portable standard, so the same MCP server can be adapted to any AI platform that supports MCP.
Prerequisites
A Totara site running version 19.1.0 or later (the Report Builder action used in this guide is available from this version onwards)
A Workato account
An External API client already set up and configured in Workato for the Totara site. Setting up this API client is not covered in this guide
Setting up Workato
Prepare a report in Totara
This example uses a Totara report to supply compliance data to the AI agent.
In the Totara site, navigate to Reports and select Create report
Select the report source needed (for example, Certification Completion) and select Create and edit
On the Columns tab, add a User ID column. This allows the AI agent to reference the user in any follow-up actions
Select Save changes
On the Access tab, under Roles with permission to view this report, ensure the role used by the API client (for example, API User) is selected, then select Save changes
Select view next to the report name to confirm the report displays correctly
If the AI agent only needs a subset of the data, such as non-compliant users, set the required filters under Search by, then select Save this search
When saving the search, set Availability to Shared, so the API client can use it
Once saved, note down the report ID and (if used) the saved search ID. Both appear in the URL of the report, for example:
/totara/reportbuilder/report.php?id=11&sid=2
In this example, the report ID (id) is 11 and the saved search ID (sid) is 2. These will be needed when configuring the MCP tool in Workato.
Create the MCP server
In Workato, create a new project using Create project
Within the project, select Create, then select MCP server
Enter the MCP server's details:
MCP server name: a name for the server (for example, "Totara Compliance")
Server instruction: a description of the server's purpose. This is used by the LLM to understand what the server is for, and is not shown to end users
Location: where the server should be created within the project
Tool source: select Project assets to add tools from within the current project
Tools can be added later, so this step can be completed without any tools selected
Select Create MCP server
Once created, the MCP server's overview page shows a Remote MCP URL. Select Copy URL to copy it, and keep it available for later, when connecting the server to an AI agent.
Configure authentication
On the MCP server, navigate to the Settings tab
Under Access Method, select an authentication method for the MCP server
Note: Token-based access is sufficient for testing, but is not intended for production. For production use, it is recommended to enable Workato's MCP verified user access, which provides secure, per-user authentication.
If using token-based access, copy the developer MCP token shown on this page using Copy
Creating an MCP tool
Create a recipe
Within the Workato project, select Create, then select Recipe
Give the recipe a name. This name will become the name of the tool available to the AI agent
Select a Location for the recipe within the project
Select Build Recipe Function as the starting point, then select Start building
Define the tool's input and output fields
The recipe function's input and output fields tell the AI agent what to provide when calling the tool, and what to expect in response.
Under Parameters schema (the tool's input), select add fields manually
Define the input fields the AI agent will need to provide. For example, a boolean field could be added to let the AI agent request only non-compliant users, with a Name, Label, Data type, and a Hint describing its purpose
Under Result schema (the tool's output), select add fields manually
Define the output fields the tool will return. For example, two string fields named
columnsandrowscould be used to return the report's column headings and row data as plain text, since an AI agent can read plain text directly
Add the tool's actions
In the recipe, add an IF condition action. This is used to check the input field defined earlier, for example, whether only non-compliant users have been requested
Within the Yes branch of the IF condition, select Select an app and action, choose Totara as the app, then search for and select the Report Builder: Get action
Note: The Report Builder: Get action is available from Totara version 19.1.0 onwards.
Enter the Report ID, and, if filtering results, the Saved search ID noted down in Step 1
Under Items to retrieve, select Columns and Rows, then define the Fields to retrieve needed for each. For example, retrieving Key and Label for columns, and Data for rows
Add a further action using the Recipe function app, and select Return data from a recipe function
Map the report's columns and rows to the
columnsandrowsoutput fields defined in Step 5. This is done using formula mode on each field, which allows a value returned by a previous step (a "datapill") to be used directly in the responseRepeat the Report Builder: Get action and the Return data from a recipe function action within the No branch of the IF condition, this time without a saved search ID, and map the responses in the same way
Note: This example does not include pagination of the report data. To handle report results with a large number of rows, a page parameter could be added to the tool for the AI agent to use, or the results could be retrieved in full using a Repeat While action that continues calling the Report Builder: Get action until no further rows are returned.
Add the tool to the MCP server
Navigate to the MCP server created in Step 2
On the Tools panel, select Add
Select Add from this project, select the recipe created above, then select Add tools
Select the pencil icon next to the tool to add a Tool description. This description helps the AI agent decide when and how to use the tool, and can include extra usage guidance
Select Save
Select the … menu next to the tool and select Start tool to make it available for use
The tool is now ready to be used by an AI agent.
Using the MCP server with an AI agent
In Microsoft Copilot
Navigate to Microsoft Copilot Studio
Use the AI builder prompt to describe the agent to be built. It is recommended to specify the desired behaviour at this stage, for example: "Create a Totara agent which will always draft and confirm content, such as emails, before continuing." This behaviour can also be edited later
In the agent editor, review and edit the agent's Instructions, model, and knowledge sources as needed
Under Tools, select Add tool, then select Model Context Protocol
Complete the fields:
Server name and Server description: a name and description for the tool
Server URL: the MCP server's Remote MCP URL, copied earlier
Authentication: select API key
Type: select Query
Label: enter
wkt_token
Select Create
Create a connection for the tool by selecting Not connected, then Create new connection
Enter the token value from the Workato MCP server's Remote MCP URL. This is the text that follows
wkt_token=in the URLSelect Create, then Add and configure
Note: For development and testing, the tool's Additional details can be set to use Maker-provided credentials, so that all chats use the token supplied when the tool was created. This is not secure for production use, since anyone with access to the agent would be able to use the MCP server via this authentication. For production MCP servers, enable Workato's MCP verified user access instead.
Test the agent using the built-in test pane
Once the desired behaviour is confirmed, the agent can be published and made available in the organisation (for example, via Microsoft Teams). See Microsoft's guidance on publishing and deploying an agent in Copilot Studio for details
In ChatGPT
Navigate to ChatGPT
Open Settings, then navigate to the Apps tab
Select Create app
Complete the fields:
Name and Description: a name and description for the app
MCP Server URL: the full URL, including the token, copied from the Workato MCP server
Authentication: select No Auth
Review the risk notice shown for custom MCP servers, and confirm understanding before continuing
Select Create
Note: ChatGPT displays a warning that custom MCP servers introduce risk, since the server has not been reviewed by OpenAI. This should be considered before making an app available to others.
To make the app available to others, select the … menu on the app and select Publish
Note: Publishing the app makes it available to all users in the organisation's ChatGPT workspace.
Review the risk prompts shown, confirm the required checkboxes, and select Publish
To use the app, navigate to Apps in the sidebar, select the app, then select Connect, and confirm the connection
Select Start chat to begin a conversation using the app, or, from any chat, select +, then More, and select the app to add it to that chat
Join the Totara Community for more resources to help you get the most out of Totara.
© Copyright 2026 Totara Learning Solutions. All rights reserved.