Basic tool profile

NovaSynth: Voice and Chat Agent Testing

NovaSynth, part of Noveum, sends simulated users through configured voice and chat agents. Teams can rehearse interruptions, changes of intent and difficult conversations over SIP calls, LiveKit audio or text, then inspect the resulting evidence before changing an agent.

Visit NovaSynth

Sources checked October 27, 2026 · Public-source review; no hands-on product test

What you configure and what you inspect

Provide the agent connection, a scenario and a simulated user's goals or behavior. The official NovaSynth page describes voice and text testing with personas, interruptions and adversarial scenarios. The useful output is a conversation to inspect and evaluate, not an assurance that an agent is ready for every real caller.

Start with a scenario whose expected outcome you can describe: for example, a caller changes an appointment request halfway through. Check the conversation and the agent's resulting action, rather than relying only on an aggregate score.

Shared credits and the free limit

The Noveum pricing page lists 2,500 free credits per month shared across its metered products. NovaSynth voice testing uses 100 credits per call minute and a text conversation uses 12 credits. Evaluation work can consume additional credits. Consequently, the free balance is not 2,500 minutes or unlimited tests.

Pro is $69 per month per organization, with 10,000 monthly credits. Monthly plan credits reset each cycle; purchased add-on credits do not expire. Choose the billing cycle deliberately: the page also offers annual plans.

Where it fits, and what still needs review

Consider this workflow when repeated conversational failure cases are hard to reproduce by hand. A simple scripted assertion may be enough for a deterministic API response. For voice behavior, review turn-taking, interrupted requests, tool actions and the conditions under which the run was collected.

Use synthetic scenarios and data you are allowed to send to the service. Confirm connection permissions, retention and evaluator costs before a larger run. This profile does not certify security compliance or detection accuracy.

Before your first run

Read the official documentation, configure a small repeatable case, and verify the resulting conversation and usage. Keep a known expected outcome so a later agent change can be compared with the same case. We have not placed calls, connected an agent or tested a paid plan.