> ## 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.
# Agent Actions
> The video gallery of every UI test run. One tile per recording, bucketed by status, with click-to-play and search/filter.
## What Agent Actions Is
A project-level archive of **every UI test recording** — each tile is the video TestSprite captured while walking through one user journey. The recording itself *is* the action log: click X, type Y, navigate to Z, all captured as video.
This is UI-only. API projects don't have it (HTTP requests don't need to be filmed; their request/response bodies are the equivalent artifact).
The **Agent Actions** tab in the project sidebar, under the **Result** group. It sits alongside the **Test Report** tab — Test Report is the aggregated outcome view, Agent Actions is the per-recording browse view.
* **Replay a failing test** to see exactly what happened on screen
* **Hand a recording to a teammate** instead of describing the bug in chat
* **Audit recent runs** at a glance, filtered to just Failed
* **Spot-check passing tests** when something looks off in the report numbers
## How the Gallery Is Organized
Tiles are bucketed by status into three sections, in this order:
| Bucket | What's in it |
| :----------------- | :------------------------------------------------------------------------------------------- |
| Passed | Tests whose run finished and assertions held |
| Failed | Tests whose run finished and assertions didn't hold |
| Blocked | Tests that couldn't complete — running, idle, or anything that didn't resolve to Pass / Fail |
Each tile shows the recording thumbnail, the test title, a priority chip, and a status badge. Clicking the tile opens an inline player.
**Tiles need a recording to appear.** Tests that ran but didn't produce a video (rare — typically a vendor-side recording timeout) are excluded from the gallery. They're still visible in the **Integration Tests** list.
## Search and Filter
The toolbar above the gallery has:
* **Search** — matches against the test title and description (case-insensitive)
* **Status filter** — Passed / Failed / Blocked
* **Priority filter** — High / Medium / Low
Filters are independent and clearable individually. The bucket sections re-render to show only matching tiles.
## What You'll See When the Project Is Mid-Run
Two transitional states:
First run hasn't produced any recordings. You'll see the in-progress placeholder with stages: **Running agents → Recording sessions → Processing videos**. Tiles trickle in as agents finish.
Recordings appear shortly after each test finishes. Tiles may trickle in over the first few seconds; the buckets reflect the final state once every test has resolved.
## Agent Actions vs. Step-by-Step
Both are visual replay tools, but at different scopes:
| | Agent Actions | Step-by-Step Walkthrough |
| :------------------- | :------------------------------------------ | :---------------------------------------------- |
| **Scope** | Project — every UI recording in one gallery | Per-test — the run of a single test |
| **Primary artifact** | Video tile | Per-step screenshots + recorded video |
| **Surface area** | Browse / filter / triage at scale | Drill into one test's flow + assertions |
| **When to use** | "Show me everything that failed" | "What exactly happened on step 4 of this test?" |
Both pull from the same recordings — Agent Actions is the index, Step-by-Step is the inspector.
Per-step screenshots, video, and behavior trace for a single test
## Edge Cases & Troubleshooting
The recording wasn't captured for this run — rare, typically a transient hiccup. Rerun the test; subsequent recordings land in the gallery normally. The original failure is still in the Test Report; the gallery just hides tiles with no playable video.
Usually a stale playback link. Refresh the page; the link is regenerated on every load.
Agent Actions is project-scoped. If your dashboard URL has multiple frontend projects under one parent (rare — typically a multi-URL setup), tiles show a project-name subtitle so you can tell them apart.
No. Recordings are static artifacts; playback doesn't trigger any agent activity. Only re-running the test consumes credits.
## Where to Go Next
Drill into a single test's run
All project-level tabs, including this one
Trigger a fresh run when a recording shows the wrong behavior
Iterate on a test in natural language after watching its recording