# Hello World: Pipecat + Coval for Voice AI Testing

May 8, 2025

## About Pipecat

[**Pipecat Cloud**](https://www.pipecat.ai/) is a fully managed platform for deploying and scaling AI agents into production — purpose-built for the Pipecat framework. It simplifies launching agentic workflows and pipelines, with features tailored for modern **Voice AI** use cases:

- Native support for agentic behavior and the full Pipecat ecosystem
- Flexible deployment options and real-time observability
- Blazing-fast performance, tuned for scale
- Cross-platform SDKs for frictionless application development
- Prebuilt templates and first-party integrations
- Enterprise-grade security and compliance

Developed by [**Daily**](https://www.daily.co/), Pipecat Cloud is backed by over a decade of infrastructure and real-time system experience.

> ✅ **New integration:** You can now connect your Pipecat agent directly to **Coval** — and run **automated Voice AI evaluations** and **quality assurance tests** with just a few clicks.

## About Coval

**Coval** is the reliability infrastructure for conversational and Voice AI agents. Whether you’re building, testing, or monitoring an agent, Coval helps you **evaluate performance, detect failures, and optimize conversation quality** at every stage of the lifecycle.

Our platform is purpose-built for **Voice AI testing and QA**, with robust simulation, monitoring, and metric tooling.

Coval is designed for:

🧪 **Developers & QA**: Automate testing across hundreds of real-world voice scenarios, IVRs, accents, and interruptions

📊 **Enterprise Agent Ops**: Track voice agent performance across vendors and standardize quality benchmarks

⚙️ **Fast-moving teams**: Catch issues before they hit production with CI/CD integration and simulation loops

## 🔧 How to Connect Your Pipecat Agent to Coval

First, Create your agent by following [Pipecat’s Quickstart guide](https://docs.pipecat.daily.co/quickstart).

Once deployed, confirm your agent is listed in the Pipecat Agent Dashboard:

Next, head over to [app.coval.dev/coval/agents/create](https://app.coval.dev/coval/agents/create) and set up your Pipecat agent. Add your agent name, details, and make sure to add your Pipecat Cloud API key.

**Note: The agent name must match the name used in your Pipecat deployment**

That’s it! You’re good to go to launch your first evaluation 💪

## 🧪 Creating Evals for Your Pipecat Agent

Use Coval to launch simulated conversations with your Pipecat agent, and get rid of the need to manually call it every time your perform a change.

Coval lets you simulate conversations with your Pipecat agent and run repeatable Voice AI evaluations every time you make a change — no more manual testing.

**1\. Define what you’re testing — Test Sets**  
Test Sets let you simulate scenarios your agent should handle, like:

• “I want a refund”  
• “Reschedule my doctor’s appointment”  
• “Navigate to account settings via IVR”

You can script everything from basic flows to multi-turn, nuanced interactions.

✅ Pro Tip: Import real production transcripts for regression testing.

**2\. Define who we’re simulating — Personas**  
Personas allow you to test Voice AI agents against realistic, human-like behavior. Customize:

• Language, accent, voice style  
• Emotional tone or background noise  
• Interruption sensitivity and speech tempo

Example: Simulate a fast-speaking, Spanish-speaking customer who interrupts often — and test how your Voice AI handles it.

**3\. Define how to measure success — Metrics**  
Metrics are the foundation of high-quality Voice AI QA.

Coval provides:

- Built-in metrics (e.g. latency, interruptions, speech tempo for audio evals)
- Custom LLM-as-a-Judge metrics (e.g., Did the agent resolve the issue? Did it follow instructions? Was it repetitive?)

🎯 Run these metrics across both simulated and production calls — creating a continuous Voice AI evaluation loop that helps you spot regressions, track improvements, and ship confidently.

## 🚀 Ready to Level Up Your Pipecat Agent?

By connecting Pipecat to Coval, you unlock a powerful workflow for **Voice AI testing, evaluation, and QA**:

✅ Simulate real user conversations before deploying  
✅ Run continuous metrics on **live production voice calls**  
✅ Detect performance drift and regressions automatically  
✅ Use consistent evaluation criteria across all environments

Whether you’re refining a new agent or managing complex customer flows, Coval helps you **build reliable Voice AI experiences at scale**.

🔗 **Try it now**: Reach out to the Coval [team](http://bit.ly/coval-demo) — we’re here to support your evaluation loop, from first simulation to full-scale monitoring.