Use Tosca MCP Server

We added this functionality in a patch. To use Model Context Protocol (MCP) with Tosca, you need to install patch 1 or later.

With the Model Context Protocol (MCP) integration, AI assistants can access your Tosca workspace and help you create, organize, update, use, and run test assets.

In this topic, you'll learn how to use the Tosca MCP Server.

Get started

To start using the Tosca MCP Server, follow these steps:

  1. Open a Tosca workspace. This workspace must stay open while you use the AI assistant with Tosca.

  2. Open your AI assistant and send a prompt.

The AI assistant then accesses the workspace with the credentials of the user of this Tosca workspace.

Please note that standard check-in/check-out rules apply, even for the AI assistant.

Available actions

Your AI assistant can perform all tasks that the Tosca Commander interface supports.

To see which tools are available, ask the AI assistant directly for the list. For example, use the following prompt:

Give me a list of all available tools

For the best results, however, we recommend that you use the Tosca MCP Server together with the skills we've created. You can get these skills from this Tricentis GitHub repository (opens in new tab).

The Data Integrity MCP toolset is an end-to-end assistant for testing the validity of your table datasets and interpreting the test results.

These are the tools the Data Integrity MCP offers:

Tool

Description

Prompt examples

Create new Data Integrity connections

The AI creates a Data Integrity connection to the database you specify.

For security purposes, the AI cannot set user passwords or connection strings. The user needs to add those manually, after AI creates the connection in Tosca.

Additionally, every Data Integrity connection that the AI creates has the tag Production. The AI can't retrieve metadata from a connection with the tag Production, which means that it can't create or fix SQL statements in Tosca for that connection. If you want to manage these SQL statements through prompts, you need to manually update the tag to Development.

Connection configurations such as Password, Connection string and Tag are in Configurations > Database > <Connection_Name>. You can also ask the AI to guide you to find them.

Note: We recommend that you use user credentials with read-only access to your database. Especially when working with connections with the tag Development.

Connection tag

Create a DI connection to my Azure database with the address tcp:myserver.database.windows.net,1433.

Create a new SQL query

The AI creates an SQL query that retrieves the data you specify from a Data Integrity connection with the tag Development. The AI can read instructions both in natural language and in pseudo-code.

For this test, include only the customer names and last order dates for shipped orders in 2026.

If order.status = 'SHIPPED' AND YEAR in last.order.on = 2026 THEN SELECT name and last.order.on.

Validate an existing SQL query

The AI verifies if an SQL query in Tosca fetches the data you want from a Data Integrity connection with the tag Development.

Does the query to my Azure database in the test "Data comparison 1" fetch the shipped orders from the last three months?

Transfer data from an SAP database to the caching database

The AI creates a test step with the Load Data into Caching Database from Customization Module and the SAP Custom Data Reader. That test step creates a table in the caching database with the data it fetches from the SAP database, according to your specifications.

I want to check the data in the Orders table in the sap-server.acme.com, instance number 01 and client 101. Can you create a test for that? 

Check specific table cells

The AI creates a test case using one or more DB Expert Modules according to your specifications. It can use an existing connection to access the table you want to check, or create a new one if an applicable one doesn't exist.

Confirm that the last cell in the first column of the table Orders in my SAP database has the value "Shipped".

Confirm that the cells in the seventh row in the columns Order.price and base.price in the table Orders in my Azure database are less than 2000.

Check the validity of a table dataset

The AI creates a test using the Complete Row by Row Comparison Module.

This tool can use your input to set up the following assets and test steps:

  • Create a Data Integrity connection and/or data transfer from SAP to access the data for the test.

  • Create an SQL query and add it to the test.

  • Set up the column mapping.

  • Set DateTime tolerances.

  • Recommend and set a row key.

This tool can also set up tests using a CSV mapping file from a data governance tool, for example: Collibra. This file describes the relationship between the two datasets you want to compare. This way, you only need to tell the AI what you want to check after you provide the file.

Check if the customer name and date of last order columns in the Orders tables in my Azure and SAP databases match.

Validate existing tests

The AI confirms if an existing test works as you describe.

I want to check if the columns with the customer names in the Orders tables in my Oracle and Snowflake databases match. But the target column is named "Customer" and the source column is named 'names'. Is the test 'Names Check' designed match the correct columns?

Update existing tests

The AI makes any change that you ask in existing Data Integrity test assets.

In the "Names Check" test, change the value of the row key to the order.code column.

Discuss test results

The AI explains the results of dataset comparison to you. It groups errors according to their type, for example: unmapped columns, DateTime mismatch, or actual data differences, and presents them to you in conversation. This way, it can let you know if it was an incorrect test configuration that caused the errors.

Check how many of the mismatches in the time column, are differences of less than a second.

Fix tests using test results

The AI suggests and apply fixes to dataset comparison tests that have incorrect configurations, after checking their test results.

AI response: Your source column uses European date format (DD.MM.YYYY) while the target order.dates column uses US format (MM/DD/YYYY) — that mismatch is what's driving all the comparison failures. I can resolve this by configuring the DateTime tolerance settings in your test. Want me to go ahead?

Note that you don't need to describe each step in test creation through prompts. A straightforward prompt will start the process. For example:

Check that the customer names and last order dates in the Orders tables in my Azure and Snowflake databases are the same.

The AI will then ask you follow-up questions for any further details it needs.

Limitations

The Tosca MCP Server provides access to many Tosca capabilities, but some workflows aren't fully supported yet.

Keep the following limitations in mind:

  • Parallel work. You and your AI assistant can work in the same workspace, but not at the same time.

  • Tosca XScan. You can trigger the scan with the AI assistant, but must perform the scan in the Tosca XScan interface. We're working to improve this experience in future versions.

  • ScratchBook. The execute_test_suite tool triggers a trial run in the ScratchBook, but currently supports test cases only.

Effective prompt patterns

If you're new to AI prompts, check out these patterns. They'll help you structure your prompts for better results.

Start with what you want to know, add the relevant background, then clearly state what you want the AI to do. This helps the AI give answers that are accurate and tailored to your situation.

[Question] + [Context] + [Desired Action]

Example:

What’s causing the failed steps [Question]
                        in the "Checkout Flow" test case from this week’s regression run? [Context]
                    Analyze the results and suggest how to fix the errors. [Action]

Ask the AI to compare two or more options, by the criteria that matter most to you. This makes it easier to evaluate trade-offs and choose the best fit.

Compare [Item A] with [Item B], focusing on [Specific Metrics]

Example:

Compare the "User Registration" and "Login" test cases, [Item A/B]
                        focusing on the number of failed steps and execution time. [Specific Metrics]
                    Highlight which one shows more stability across recent runs.

Describe the problem, provide the necessary context, and ask the AI to investigate possible causes or solutions. The more specific you are, the better the troubleshooting will be.

[Problem Statement] + [Context] + [Investigation Request]

Example:

The "Payment Validation" Module is missing from the Regression folder. [Problem Statement]
                        It should be under "Release1". [Context]
                    Find out where it was moved and put it back into the correct folder. [Investigation Request]

Specify the type of report you need, who it's for, and what it should include. This helps the AI produce content with the right structure, level of detail, and tone.

Create [Report Type] for [Audience] including [Specific Elements]

Example:

Create a summary report for our QA team including:
                        - The latest results
                        - The number of passed and failed test cases
                        - The top three recurring test failures
                    - Recommendations for improvements