> For a complete page index, fetch https://docs.synthflow.ai/llms.txt. For full documentation content, fetch https://docs.synthflow.ai/llms-full.txt.
# Manual Testing
> Learn how to validate your voice agent before launch with phone, chat, widget, and simulation-based testing.
Before you deploy an agent to production, validate how it behaves across real conversations, edge cases, and channel-specific experiences. Synthflow gives you four testing methods that support both quick iteration and repeatable pre-release validation.
Open the agent editor and click **Test** to choose **Phone call**, **Chat**, **Widget**, or **Simulation**. Use manual testing for fast spot checks, then use [Simulations](/simulations) to run structured suites before launch.
## Choose a testing method
You can access all testing options from the **Test** menu in the agent editor.
### Phone call
Use **Phone call** to test with a real call to your number. Click **Call Me** to hear how your agent sounds in a live interaction and verify pacing, latency, and call quality.
Choose this method when voice quality and conversational realism are your priority.
### Chat
Use **Chat** to validate conversation logic quickly with text input. The chat panel helps you inspect turn-by-turn responses and iterate on prompts without placing calls.
Choose this method when you want fast debugging loops for prompt behavior.
### Widget
Use **Widget** to test the browser-based voice experience that end users will see on your site. Start a conversation directly in the embedded widget and verify client-side behavior.
Choose this method when you need to validate the web delivery channel.
### Simulations
Use **Simulation** to run automated test suites across many scenarios in parallel. Simulations pair your agent with a persona agent (a simulated customer) and score outcomes against your success criteria.
Choose this method for systematic validation, regression prevention, and release readiness. For full setup details, see [Simulations](/simulations).
## When to use each method
**Manual testing** (Phone call, Chat, Widget) is best for fast checks while you build. It lets you inspect one conversation at a time and quickly diagnose specific issues.
**Simulations** are best for pre-release validation. They let you test many scenarios in parallel with measurable criteria so you can catch regressions before production.
**Recommended workflow**: Iterate quickly with Chat, confirm channel behavior with Phone call or Widget, then run a full simulation suite before deployment. Re-run simulations after major prompt or workflow changes.
## Why run simulations before release
Manual testing can take hours when you need broad coverage, and edge cases are hard to reproduce consistently. Simulations automate that coverage so validation is faster and more reliable.
With simulations you can:
* Validate agent behavior across multiple scenarios simultaneously
* Test edge cases and failure scenarios that are hard to trigger manually
* Iterate faster when refining prompts, testing changes in minutes instead of hours
* Ensure consistency before client deployments with repeatable test suites
* Track improvements over time by comparing results across test runs
Create automated test suites, define success criteria, and validate your agent at scale.
## FAQ
Start with manual testing while you build, usually with **Chat** for speed and **Phone call** for voice quality. Before release, run simulations to validate broad scenario coverage and catch regressions.
Run simulations before every production release and after major prompt, tool, or workflow updates. For high-volume agents, schedule regular simulation runs to monitor quality over time.
No. Simulations provide scalable, repeatable coverage, but manual testing is still valuable for listening to real call quality and checking user experience details in chat or widget channels.