> ## Documentation Index > Fetch the complete documentation index at: https://docs.testsprite.com/llms.txt > Use this file to discover all available pages before exploring further. # Feature Exploration > How TestSprite explores your live app feature by feature: what you see, where the results live, and how to re-run exploration. An AI agent exploring the app under test: live preview on the left, the steps it took on the right ## Overview Before writing any UI tests, TestSprite's testing agent visits your live app and explores it feature by feature: opening real pages, trying common user flows, signing in if you [configured login](/web-portal/core/ui/ui-authentication), and recording what it could and couldn't reach. The test plan is then grounded in **how your app actually behaves**, not just how your PRD describes it. Feature exploration runs only for UI projects. API projects discover endpoints from your docs and probe them against your live API instead; see [API Discovery](/web-portal/core/api/api-discovery). Some features may explore only partially. You can still continue to the test plan: TestSprite uses what it observed and falls back to your PRD or description for the rest. ## How It Works The agent opens your starting URL, identifies your features (e.g. *Sign Up*, *Dashboard*, *Checkout*), and walks through each one, clicking buttons, filling forms, and moving between pages. Features are explored in parallel, so a long flow doesn't hold up the rest. Plans written from what it observed are more reliable: * **Steps reference real selectors and copy** from your pages * **Edge cases come from real screens**, not assumed ones * **Auth flows match your actual login**, including unexpected redirects * **Features behind paywalls or plan gates** in your app are flagged, instead of producing tests that can't run Exploration starts as soon as you click Create project and typically takes about **3 minutes** for an average app; longer flows or large feature sets take more. You can leave and come back: if it's still running when you return, TestSprite picks up where it left off. ## What You'll See During Exploration While exploration runs, the project's Sources page shows a live view headed **AI agent exploring your app** (or **AI agents exploring your app** when several are working): The exploration view with Live Explore, Steps, Page Visited and Captured Pages labelled | Element | What it shows | | :- | :- | | Progress counter | With several agents, how many have finished (e.g. *"2/4 agents finished"*) | | Agent status | Each agent's state: Starting, Logging in, Exploring, then Done or Failed | | Live browser preview | Streams TestSprite navigating your app in real time | When exploration finishes, TestSprite composes your coverage map on Test Coverage. ## Reviewing the Results Sources page: Live Site Exploration with its status chip, recording, and Re-explore Results live under **Live Site Exploration** on the project's Sources page. The **Explore your site** card has a status chip (Processing while agents work, Ready when done, Failed if the run didn't complete) and one card per exploration run with its recording and captured pages. Click a run to open its recording and report. ## Adjusting Before Plan Generation If TestSprite couldn't sign in or reached a different URL than expected, open Project Settings → Test Execution → Environment and correct the environment's URL or test account. Back on Sources, click Re-explore on the **Explore your site** card. ## Tips for a Good First Run If your app requires sign-in, share a test account when you create the project. Without credentials, exploration is limited to what's accessible without login. Use the URL a real user lands on (your homepage or app dashboard), not a deep link into one tool. TestSprite uses the starting URL to discover the rest. Add hints like *"Skip the help center"*, *"Focus on the checkout flow"*, or *"The dashboard route is `/app/v2/home`"*. TestSprite uses them to guide both exploration and plan generation. ## Where to Go Next The full UI testing flow, with exploration as one step What happens with the results next This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.