Skip to content
Semaloop Docs

Recommended setup

Once builds are arriving, a few things make the difference between tests that run and tests your team relies on. Work through them in this order.

  1. Connect Slack. Each build’s results land in a channel your team already reads, so nobody has to remember to check. See Slack.
  2. Add test accounts as secrets. Give the agent accounts to sign in with, and enough of them for tests to run at the same time without sharing one. See secrets and credentials.
  3. Add a custom tool to reset or create users. Starting every run from a known state is the biggest single fix for flaky tests. See custom tools and sign-in and user state.
  4. Write a handful of tests for your most important flows. Sign-up, sign-in, purchase and your core feature are a good start. Describe the goal rather than the taps; see best practices.
  5. Tag your tests and choose what runs on each build. For example, run everything on release candidates and only a smoke set on every staging build. See tags and automatic runs.
  6. Connect your coding agent to the MCP server. It can write and fix tests, and investigate failures with the run’s logs and video. See MCP server.
  7. Triage issues and leave feedback. Resolve or archive the issues Semaloop finds, and use thumbs up and down on results, so what’s flagged stays useful. See managing issues.