Generating test scenarios...

Running simulations...

Grading results across 25 metrics...

0s elapsed

0 --

Your Agent Grade

Based on 5 simulated conversations scored across 25 metrics

Dimension Scores

Top Issues Found

Want to fix these issues? Get unlimited testing, auto-fix recommendations, and production monitoring.

Start Free Trial

Something went wrong

We couldn't grade your agent right now. Please try again.

What You'll Get

A Full Quality Report in 60 Seconds

No more guessing if your agent is ready for production. Get a data-driven grade across the metrics that matter.

A+

Overall Grade

A letter grade (A+ to F) and numeric score (0-100) based on 25 weighted metrics across 6 dimensions.

Score: 87 / 100 — Grade: B+

Top 3 Issues

The highest-impact failures your agent has right now — what's breaking and why, ranked by severity.

Hallucination in pricing scenario

✓

Fix Recommendations

Actionable steps to improve your score. Specific prompt edits and guardrails you can implement today.

Add price confirmation loop

How It Works

Three Steps. Sixty Seconds.

Paste Your Prompt

Copy your agent's system prompt and paste it into the form above. Works with any platform.

We Simulate 5 Conversations

Our engine runs greeting, objection, hallucination, silence, and compliance scenarios against your prompt.

You Get a Detailed Grade

Receive an overall score, dimension breakdown, top issues, and specific fix recommendations.

FAQ

Common Questions

Is this really free? ▾

Yes, 1 free grade per agent. No credit card, no signup. We send the full report to your email. Unlimited testing, production monitoring, and auto-fix recommendations start at $49/mo.

What gets tested? ▾

5 core scenarios covering greeting quality, objection handling, hallucination detection, silence recovery, and compliance. Each scenario is scored across 25 metrics in 6 dimensions: Conversation (30%), Task Completion (25%), Safety (15%), Empathy (10%), Latency (10%), Audio (10%).

How accurate is it? ▾

We use GPT-4.1 for grading with a calibrated 25-metric rubric tested against human evaluators. Our scoring engine achieves 94% correlation with expert human graders across a golden dataset of 200+ agent transcripts.

Do you store my prompt? ▾

Only to generate your grade. We don't use your prompt for training, share it with third parties, or retain it after grading is complete. Your data is encrypted in transit and at rest.

Which platforms are supported? ▾

VoxGrade works with any voice AI platform that uses a system prompt. We have native integrations with Retell AI, Vapi, ElevenLabs, LiveKit, and Bland AI, but any prompt-based agent can be tested.

Grade Your Voice Agent — Free