> ## Documentation Index
> Fetch the complete documentation index at: https://docs.twill.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Live Previews for Review

> Twill self-verifies against a running app, and you can open previews to validate changes before approving PRs

Every task runs in its own [fork of the workspace sandbox](/sandbox-environment) — a real dev environment with your repo, dependencies, and dev servers. Twill doesn't just edit files — it can boot your project, run checks, and validate changes against a running app. Any port the agent opens in the fork is exposed as a public preview URL, and you can SSH into the same fork from the task page to review what changed and approve with confidence.

## How Twill Uses Live Previews to Self-Verify

Depending on the task and your repo, Twill may:

* **Start your app** via the repository entrypoint script
* **Run tests / lint / typecheck / builds** to catch regressions
* **Exercise UI flows** with browser automation when the change is user-facing
* **Inspect logs** to confirm services are healthy and errors are not introduced

## Web Preview

Click **Preview** to open your running application on any exposed port. Each port gets a unique public URL that works in your browser — useful for quickly validating UI and behavior during review.

## SSH Access

Connect your local IDE (VS Code, Cursor) directly to the sandbox via SSH. This lets you browse files, run commands, and debug as if you were working locally. SSH and IDE-remote work for both the workspace sandbox and any task fork — see [Fixing your dev environment](/sandbox-environment#fixing-your-dev-environment) for when to use which.

## Logs

Monitor your services in real-time. Each service that writes to `$TWILL_ENTRYPOINT_LOG_DIR` appears as a separate tab:

* **dev-server.log** — Frontend output
* **api.log** — Backend output
* **entrypoint.log** — Startup script output

Logs stream live as the agent works, so you can watch exactly what's happening.