Grades how easily coding agents can start with a product, and records the sessions where they stall
NovaSynth by Noveum
New· 6Simulates synthetic callers and chat users against voice and chat agents, then scores what breaks in each call
About NovaSynth by Noveum
Voice agents tend to fail in ways a script of tidy test sentences never reveals: a caller interrupts, a line is noisy, someone switches language halfway through a sentence. NovaSynth, released by Noveum on 17 September 2026, tests voice and chat agents by reproducing those failures before real customers meet them.
Synthetic callers with goals and moods
NovaSynth generates synthetic users, each with a goal, a mood, an accent and a pattern of behavior. Examples from the maker include a frustrated caller on a third contact about a refund, a rapid interrupter who reverses a decision halfway through a sentence, a caller who changes language, and someone trying to coax the agent into disclosing its hidden instructions. Personas and scenarios are drawn from the agent's own context, meaning its prompt, requirements and documentation, instead of being written by hand, and they can be run as a full matrix of personas against scenarios in one batch.
Real phone calls rather than mocked transcripts
The calls travel over real telephony. Calls go out over SIP using Twilio, Plivo or Vobiz, or run on LiveKit real-time audio. Integrations also cover Pipecat, LangChain, LangGraph, the SDKs of OpenAI and Anthropic, and plain chat over HTTP. During a run, lookups the agent makes against its tools are replayed, while anything it tries to write is kept in a sandbox. A changed decision can therefore be tested all the way through without touching production systems.
Scoring and load
Each call is captured as a trace and scored, audio included. Noveum describes over 30 voice metrics in this scoring, and over 100 on its wider platform. For load testing, concurrency can be raised to 2,000 lines.
How Noveum charges
Noveum prices per organization rather than per seat, using a shared credit system with a free tier and several paid tiers, and annual billing is cheaper than monthly. Its own tagline for the product is about breaking agents before customers do.
TrustList summary
NovaSynth by Noveum is a testing product for voice and chat agents, released on 17 September 2026. It builds synthetic users, each with goals, moods, accents and behaviours, and runs them through the agent under test. The calls are real telephony rather than mocked transcripts, so a failure shows up the way a customer would meet it.
Features
- Synthetic callers: generated users such as a frustrated repeat caller, someone who talks over the agent, or a person who switches language mid-sentence.
- Live telephony: calls agents by SIP through Twilio, Plivo or Vobiz, and streams audio over LiveKit.
- Scenario matrix: runs a full set of personas against a set of scenarios under one batch.
- Safe replay: replays read calls to the agent's tools and sandboxes writes, so production systems are not touched.
- Scoring: records each call as a trace and scores it, including the audio, against more than 30 voice metrics.
Clients
Teams building voice or chat agents that handle calls or requests from customers, including those who want load testing with up to 2,000 concurrent lines.
Key facts
- Frameworks: Pipecat, LangChain, LangGraph, OpenAI and Anthropic SDK code, and HTTP chat.
- Pricing model: billed per organisation on a shared credit system.
- Release: 17 September 2026.
Summary written by TrustList from the business's own published information, Nov 2026.
NovaSynth by Noveum launches
Categories & features
Trust Score
An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.
Updated 9/18/2026
Request a demo or quote from NovaSynth by Noveum
Reviews
Write the first review of NovaSynth by Noveum
Used it? Your experience helps other buyers decide.
Write a reviewQuestions & answers
No questions yet. Be the first to ask about NovaSynth by Noveum.
NovaSynth by Noveum alternatives
Similar listings buyers compare against this one.
Browser testing from the command line: record a flow once, replay it headless, and collect backend logs for the run
Mobile app testing that a coding agent can run on a real device and read back as a report
Automated checks that continuously find breakages in a product, for engineering teams
Tests Model Context Protocol servers inside AI clients like ChatGPT, Claude and Copilot, and blocks regressions in CI
Mac app that tests multi-user features with AI participants, each with its own login, browser and goal
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More Automated Testing SoftwareBack to the ranked category
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.