Today we’re introducing the Stb-tester Agentic API: an MCP (Model Context Protocol) server that lets AI coding assistants — Claude Code, Cursor, GitHub Copilot, and others — develop and debug your Page Objects for you. It closes the loop on page object development and debugging and will become more and more powerful over time, as we expose more of our features through the Agentic API and as AI assistants become more capable.

Developing and updating page objects is an iterative process: you write the simplest possible code, run it against captured frames in the Object Repository, check the output, and fix it if it’s wrong. Later if you discover a case where it doesn’t work or the UI changes you save that frame to the object repository and repeat the process.

AI agents have for a long time been able to write reasonable code, but until now they couldn’t use the object repository to check their work against real frames. With the Agentic API, they can, closing the agentic development loop and letting them do the work for you.

This enables them to:

• Write a Page Object class for you.
• Debug and fix page objects that are returning the wrong values.
• Understand the behaviour of the page objects so they can use them in test scripts and even write test scripts for you.

The result is a Page Object that has already been verified against real captured frames, rather than plausible-looking code for you to debug by hand.

Setup takes about a minute — generate an access token, run stbt_rig setup --vscode, and point your assistant at your test-pack. See the Agentic API manual page for the details, including Claude Code and VS Code.

Bring your own agent

The Agentic API is “bring your own agent”, and we’ve designed it with privacy first:

• We don’t run agents, and we don’t use third-party AI models. There is no “Stb-tester AI” acting on your behalf. All of the agentic intelligence comes from the assistant you bring; the Portal only executes the specific tool calls it receives.

• We don’t send your test code to a third party. Your assistant reads your code locally, and what it forwards to its own AI provider is governed entirely by your choice of assistant — not by us.

• Everything runs as you. The MCP server uses your access token and requires push permission, exactly like the REST API. Your assistant can only do things you could do yourself, and you can revoke the token at any time.

Because you bring the agent, you can even pair it with a privately-hosted model so that no code or data need leave your control at all.

This is the first feature exposed through the Agentic API, and we’ll be adding more over time. Try it out, and send your feedback to support@stb-tester.com.