Introducing Copilot and local MCP server

Why we built this

End-to-end tests shouldn’t slow you down. But products change and maintenance piles up.

Teams want the convenience of asking for a test in plain English and seeing it appear in the editor, already runnable and easy to tweak. They also want their existing AI coding tools to be able to iterate on tests with full browser context.

Momentic Copilot and MCP server make that workflow real. You stay in flow, in your IDE, describe what you want, and Momentic handles the rest.

What we built

Copilot in your editor

Open the Copilot panel, describe the flow you want, and watch a runnable test appear. Ask follow‑ups like “add an assertion that the chart is visible” or “refactor steps into reusable modules.”

Copilot looks at the current page, inspects your modules, edits the test, and runs steps so you can validate changes quickly.

Local MCP server

For teams using AI coding tools, the local MCP server exposes Momentic’s testing agent over SSE. Point your IDE to the local endpoint and give your AI assistant first‑class access to your tests, modules, environments, and a live browser session.

We're finally closing the feedback loop for coding agents. Time to make sure what the AI spits out actually works.

Works with Cursor, Windsurf, Claude Code, and others that speak MCP.

{
  "mcpServers": {
    "momentic": {
      "serverUrl": "http://localhost:58888/sse"
    }
  }
}

How it works

Under the hood, Copilot and the MCP server use a set of tools to observe the app, edit tests, and run them against a local live browser session.

Project

• list_modules to list all modules in the active project
• list_tests to list all tests in the active project
• get_test to get the contents of a test
• get_module to get the contents of a module

Browser

• observe_page to look at the current page

Edit

• create_test to create a brand new test based on a description
• edit_test to edit a test using Momentic’s testing agent
• add_step to add a step
• edit_step to edit or replace a step
• remove_step to remove a step
• splice_steps to splice steps into the test
• overwrite_steps to completely replace the test

Session

• run_step to run a step that is in the test
• reset_session to reset the editor state and browser session
• get_environment_variables to read current environment variables

Preflight

• preview_preflight to validate a locator before you commit
• preview_step to dry‑run a step without adding it to to the test
• preview_ensure to verify a condition is true in the browser

All of these are the same capabilities you have in the Momentic app, now exposed where you live: your IDE.

Real use cases

Copilot and MCP unlocks a 10x productivity boost.

Even though our low-code editor is already leagues ahead of building Selenium scripts yourself, there’s friction, albeit small, to start up the local app and use our editor to iterate on tests.

Here are new workflows that are possible:

1. Build new tests: “Build a test for the new share feature”
2. Update existing tests: “The placeholder for a new page is now ‘New page’, can you update all of the steps that reference the old one?”
3. Migrate existing tests: “Migrate the editor Cypress tests to Momentic”

Building the AI-native testing platform

Momentic serves the entire engineering org from one shared source of truth for tests.

Engineers create, edit, and run E2E tests directly from the IDE with Copilot. QA and SRE see analytics and results before a PR merges, so bugs get caught early. Product teams catch UX regressions from real user flows.

AI continuously observes your app, keep tests updated, and verifies that it’s actually working as intended.

What’s next

Now that Copilot and the local MCP server bring Momentic into your local editor and AI coding tools, we’re building even deeper integrations and more intelligent AI to keep tests always up to date.

If you’re ready to turn plain English into reliable coverage, book a demo to get started with Momentic.