Improvement
We've improved the reliability of AI-generated tests. You should see fewer failed generations, more consistent step sequences, and better results when testing sites with dynamic content.
Improved reliability for AI Search (Fallbacks) and Assertions
Improvement
We've made two updates to AI Search & Assertions to improve predictability and reliability.
Upgraded AI model. We've switched to a newer underlying model with improved visual recognition, particularly for ambiguous elements and icons without text labels. Most tests should see more consistent results without any changes on your end.
Exact matching for quoted text. When quotes are used in an AI assertion, the model will now look for an exact match.
Rainforest MCP Server: Generate and Run Tests Right From Your Editor
New
You can now connect Rainforest to AI-powered coding tools like Claude Code, Cursor, and Windsurf using the Model Context Protocol (MCP). Create and run tests in natural language without leaving your editor — right alongside the code you're building. Because your AI assistant already has context on your codebase, it can generate higher-quality tests that reflect your actual app flows. No more context-switching to a separate tool just to write tests. Learn how to set it up →
Create tests using a coding agent (update to Rainforest CLI)
New
You can now start generating tests directly from the command line, without needing to open the web app. This means you can use a coding agent like Claude Code, Codex, or Cursor to kick off test generation as part of your development workflow — no context-switching required. Point your agent or read more here.
Faster scroll actions in generated tests
Our computer-use agent, the agent that interacts with the virtual machine when generating a full Test or snippet, now has access to see the entire webpage and jump to that element directly. This means we can generate scroll actions significantly more quickly.
Improvement
Choose which browser/platform combination to use to generate a test with AI
Improvement
Now you can select one of three supported browser/platform combinations when creating or regeneration a test with AI.
Create an test plan for your entire application with AI test planner
New
Our new Rainforest AI Test Planner uses AI to crawl your application from a starting URL, methodically exploring your entire product. We deploy 7 agents that attempt to navigate as many accessible paths as they can discover. After this exploration is complete, we return a PDF organized into areas, features, and specific test flows that should be covered.
Learn more here.
Regenerate a test with AI
New
You can now regenerate any test with AI. Just select “Regenerate” and enter a prompt.
If the test was originally created with AI, we’ll save the previous prompt so you have a starting point. This is especially helpful if the AI got stuck the first time—simply refine the prompt and try again.
Test generation AI agents self-document with comments
Improvement
Using our test generation feature to create tests via prompt rather than writing out actions manually means that the work to create a new test shifts from writing the test to reviewing the work that AI did for you. This should be easier now that our AI agents self-document. As our agents plan their next actions while executing on your prompt, they now add comments so that each set of steps are associated with a specific task. This should make reading through the output of a flow easier.
Use snippets to log in or set up test generation
Improvement
When drafting a new test with AI, you can now use a snippet to set up that flow. When you kick off a request, we'll run through the steps from that snippet before kicking off the rest of the process (i.e., using our AI agents to work through the flow on the VM and write new steps).
This will save you time both on generation and the time you would spend editing the test that AI creates.
