Skip to content
Get Free Trial

How Atlas Works

How Sahi Pro Atlas turns the application behaviour you describe into Sahi Pro tests, and where AI is used along the way.

Sahi Pro Atlas turns the application behaviour you describe into Sahi Pro tests. This page walks through each stage, from your description to the files Sahi Pro runs.

  1. Describe the application behaviour. Type it in, upload a document, or describe it in a chat message. You can also use your own editor’s AI: it gathers the context and sends it to Atlas through the Atlas MCP server.
  2. Atlas builds a spec. AI structures the application behaviour into a mapp.yaml spec of states, roles, conditions and outcomes. It suggests where in the UI each behaviour lives, and turns fields and preconditions into typed forms. The spec is a simple format that people can read and edit.
  3. You review and approve. Every proposed change is shown as a diff. You can edit the proposed values directly, not just accept or reject them, and nothing is written to the spec until you approve it.
  4. Atlas plans the tests. Fixed rules generate one scenario for each path, error and access rule, plus every real combination of choices that affect each other. The same spec always produces the same tests.
  5. Atlas writes each test case. Each test case is structured, with its setup, teardown and seed data worked out by convention.
  6. Atlas creates the Sahi Pro files. Test case steps are mapped by rule to Sahi Pro’s own formats: scenario files (.s.csv), .sah files and data files (.csv). Existing keywords are reused, and new keywords are created as stubs, without steps.
  7. You implement new keywords in Sahi Pro. Record the interaction once by walking through the application, or let AI Suggest author the keyword for you. Sahi Pro saves the keyword file.

Ask an AI the same question twice and you can get two different answers. For tests, that’s a real problem: a test suite that changes on its own can’t be trusted.

Atlas uses AI only to understand the application behaviour and build the spec. The tests come from that spec through fixed rules, so the same spec gives the same tests, every time. When a test changes, it’s because the spec changed, and you approved that change.

Stage How it works
Describe the application behaviour, build the spec AI
Review and approve You decide
Plan the tests, write the test cases, create the Sahi Pro files Fixed rules: the same spec gives the same tests every time
Implement new keywords Your choice: record it in Sahi Pro, or use AI Suggest