Okareo

Okareo: synthetic users that break your agents before your customers do

Okareo is a testing and evaluation platform for voice and text AI agents, founded in 2023 by Matthew Wyman and Boris Selitser. Its core abstraction is the "Driver": a synthetic simulated user with a personality, goals, and context that stress-tests an agent through realistic multi-turn conversations across 30+ languages. For voice agents, Drivers reproduce the conditions that actually break production calls — background noise, crosstalk, clipping, mid-call language switches.

Evaluation combines judge-based, symbolic, and audio checks, wired into CI/CD so a failing agent blocks the pull request. A unified trace-and-transcript view supports debugging, OWASP LLM Top 10 testing covers the security dimension, and a production loop converts real failed calls into regression scenarios automatically. Developer surface includes an API, CLI, MCP access, and open-source SDKs on GitHub.

Key Features

• Synthetic Driver users simulate realistic multi-turn conversations in 30+ languages
• Voice testing under audio conditions: noise, crosstalk, clipping, language switching
• Judge-based, symbolic, and audio evals with CI/CD pull-request gating
• Production failures auto-converted into regression test scenarios
• OWASP LLM Top 10 compliance testing
• API, CLI, MCP access, and open-source SDKs (github.com/okareo-ai)

Ideal Use Case

Teams shipping customer-facing voice or chat agents — conversation designers, ML engineers, QA — who have outgrown manual spot-checking. Okareo fits the moment an agent goes from demo to production: systematic simulation before release, eval gates in CI, and a feedback loop that turns every production failure into a permanent test.

How Okareo differentiates

General LLM-eval platforms like LangSmith or Braintrust focus on text traces; Okareo's depth is voice — audio-condition simulation, multilingual switching, and telephony realism that text-first tools do not model. Pricing is transparent and self-serve (Pro $199/mo, Team $799/mo, 14-day trial without a credit card), backed by AV8 Ventures and Bee Partners, with an active open-source SDK footprint.

FAQ

What is Okareo? A platform that tests voice and text AI agents using synthetic simulated users across 30+ languages, with evals wired into CI/CD so regressions are caught before customers hit them.

How much does Okareo cost? Pro starts at $199/month and Team at $799/month, both usage-capped with per-unit overages; a 14-day trial requires no credit card, and Enterprise adds SSO, RBAC, and VPC or on-prem deployment.

Does it test voice agents specifically? Yes — audio-quality simulation (noise, crosstalk, clipping) and mid-call language switching are core features, not add-ons.

How does it fit into CI/CD? Evals run as checks that can gate pull requests, and production failures are automatically converted into regression scenarios.

tl;dr

Okareo simulates the users your agents will actually face — noisy, multilingual, unpredictable — and turns those simulations into CI-gated evals, with self-serve pricing from $199/mo.