Join our Discord
MCP servers in Test Companion
Connect Test Companion to Jira, Slack, Sentry, and other tools through MCP servers, so it reads context and takes action in those tools from chat.
Your requirements sit in Confluence, and your work items in Jira or Azure Boards. Your test cases sit in Zephyr Scale. Test Companion runs in your IDE. Without a connection between them, you paste context into the chat and carry results back by hand.
Model Context Protocol (MCP) is an open standard. It gives an AI agent one fixed way to call tools in another product. Each MCP server publishes a list of tools, such as create a work item or send a message. When a task needs one of those tools, Test Companion shows you the call. After you approve it, Test Companion continues with the result.
MCP servers are available in Visual Studio Code only. Support for JetBrains IDEs, Visual Studio, and Eclipse is in progress.
What you can do with MCP servers
After you connect a server, you describe the task in chat as usual. You do not name the tool. Test Companion picks the server and the tool when your request needs them.
| Testing task | Example prompt |
|---|---|
| Plan tests from a requirements page | Read the Confluence page "Checkout requirements" in the PAY space and generate test cases for the payment flow. |
| Start from a work item | Get Jira issue SHOP-231 and write Playwright tests for its acceptance criteria. |
| Debug a failure with production data | Find the Sentry error behind this failing checkout test, and check the Datadog logs for the same time window. |
| Record results in your test management tool | After this run, update the execution status of the matching test cases in Zephyr Scale. |
| Share results | Create a Linear issue for each new failure, then post a summary to the #qa-releases channel in Slack. |
Test Companion can chain these steps in one task. One task can read a work item, generate and run the tests, and record the results in Zephyr Scale. It can then post a summary to Slack. You approve each tool call as it happens.
MCP servers and integrations
Test Companion has two ways to work with an external tool. They answer different needs, and you can use both.
| Aspect | Integrations | MCP servers |
|---|---|---|
| Purpose | Bring data from a tool into a built-in Test Companion flow, such as test cases from a Jira issue or a Figma design | Give Test Companion tools it can call in the external product during any task |
| Setup location | The Integrations page of your BrowserStack account | The MCP Servers view in Test Companion |
| Usage | Paste a link or pick an item in a guided flow | Describe the task in chat, then approve each tool call |
| Best for | Repeatable flows with a fixed output | Open-ended tasks that span more than one tool |
Jira and Confluence are available both as an integration and as an MCP server. The two connect separately.
Supported MCP servers
Each server in the Catalog tab of the MCP Servers view is a hosted server that BrowserStack has checked. You connect it with one sign-in to your own account in that tool.
| Server | Capabilities | Sign-in |
|---|---|---|
| Atlassian | Search, create, and update Jira issues and Confluence pages. | Atlassian account, OAuth |
| Azure DevOps | Work with Azure Boards, Repos, Pipelines, and Test Plans. | Microsoft account, OAuth |
| Linear | Find, create, and update Linear issues, projects, and comments. | Linear account, OAuth |
| Zephyr Scale | Manage Zephyr Scale test cases, test cycles, and test executions. | Zephyr Scale account, OAuth |
| Sentry | Look up Sentry issues, errors, and releases. | Sentry account, OAuth |
| Datadog | Query Datadog logs, metrics, monitors, and incidents. | Datadog account, OAuth |
| Slack | Send and read messages in the channels you can access, and look up channels and users. | Slack account, OAuth |
The catalog is served by BrowserStack, and your organization can ask BrowserStack to hide a server for some users. For any other server, see Add your own MCP server to Test Companion.
How it works
Using an MCP server has three parts. You connect the server once, then ask and approve during each task:
- In the MCP Servers view, open the Catalog tab and click Connect on a server. Sign in with your own account for that tool.
- Describe your task in the Test Companion chat. When the task needs an external tool, Test Companion shows a card. The card names the server and the tool it is about to call.
- Click Approve to run the tool, or Reject to skip it. Test Companion continues the task with the result.
Security and control
You decide what Test Companion can do in each connected tool, and the tool’s own permissions still apply:
- You sign in to each server with your own account. Test Companion can reach only what that account can reach, within the access you grant at sign-in.
- By default, every tool call shows an approval card in chat. You can auto-approve tools you trust, one tool at a time or a whole server.
- You can turn off one tool, or a whole server, without signing out.
- Sign-in tokens are saved in the secure credential storage of your IDE on your machine. BrowserStack does not store them.
- Sign out deletes the saved tokens for a server. Delete removes the server from your settings.
The controls are described in Manage MCP servers in Test Companion.
Next steps
- Connect an MCP server to Test Companion: Connect a server from the catalog and approve its first tool call.
- Add your own MCP server to Test Companion: Add a server that is not in the catalog.
- Manage MCP servers in Test Companion: Read a server’s status, control its tools, and sign out.
We're sorry to hear that. Please share your feedback so we can do better
Contact our Support team for immediate help while we work on improving our docs.
We're continuously improving our docs. We'd love to know what you liked
We're sorry to hear that. Please share your feedback so we can do better
Contact our Support team for immediate help while we work on improving our docs.
We're continuously improving our docs. We'd love to know what you liked
Thank you for your valuable feedback!