// the_platform
Your enterprise, simulated.
Veris builds a high-fidelity twin of your stack and the people who use it: every API, every database, every human in the loop. Build agents, evaluate an AI system’s quality and cost, or ship code written by AI. It all meets reality here first.
// 01_actors
Actors that behave like your users.
High-fidelity personas for voice and text. Each actor has a goal, a temperament, and a history. They push back, get confused, interrupt, and change their minds, and they stay in character across an entire conversation. Your conversational agents get tested against people, not scripts.
- realistic voices, with accents and background noise
- text actors in any language your customers speak
- consistent personality, goal, tone, and fixed assets
// 02_twins
Twins of every system you depend on.
High-fidelity simulators of the dependencies your stack talks to, external and internal alike. More than 70 are ready today: CRMs, billing, ticketing, databases, queues, clouds. For any system we don’t have yet, hand Veris a spec and it builds the twin.
- stateful and consistent, not canned mocks
- simulates the logic and the failures: declines, timeouts, expired tokens
- internal services twinned from your own specs
…and dozens more, from banking cores to ERPs. Or bring an API spec and Veris builds the simulator.
// 03_scenarios
Scenarios written from your reality.
AI tools compose test scenarios from the sources you already have. Start from PRDs, policies, and regulations, from your agent’s harness and instructions, from production logs, or from a plain-language prompt. The result is hundreds of situations, each with a verifiable outcome.
- documents in: PRDs, policies, regulations
- your harness and instructions, read directly
- production traces, pulled in through an integration
- or just describe the situation in a prompt
// 04_runtime
One engine keeps every world honest.
Compose an environment from any number of twins and actors. The simulation engine keeps every component consistent with the scenario, so the database agrees with the CRM and the caller knows what the ticket says. Then run it at scale: thousands of isolated sandboxes, scheduled and orchestrated, each one you can reset, fork, and replay.
- consistency & fidelity enforced across every component
- isolated sandboxes: reset, fork, and replay any run
- every sandbox reachable over MCP
- every agent step and LLM call captured via proxy
// 05_root_cause
Every failure becomes a fix.
Runs are checked against each scenario’s expected outcome, and failures don’t stop at a red X. Veris traces each one to the instruction or tool that caused it, proposes the fix, and pins the scenario so it can never regress quietly. For regulated teams, the same evidence rolls up into an audit report against the regulation you name.
- root-cause analysis with concrete, testable fixes
- failures pinned as permanent regression scenarios
- audit reports mapped to the regulation you name