Frequently Asked Questions​

Pricing depends on how you deploy — the number of agents, patient or conversation volume, the integrations you need (EHR, wearables, Centralive), and your support and compliance requirements. Because clinical and research deployments vary widely, we scope pricing per request.

Yes — evaluation in Livi is yours to design. You can define the metrics that matter for your use case, set how each is scored, and assign the pass/fail thresholds an agent must meet before it's considered ready. Out of the box you get clinically relevant measures like Clinical Accuracy, Guardrail Compliance, Source Grounding, and Edge-Case Handling, but you're free to add, adjust, or weight criteria to fit your protocol or care standards. Evaluation is layered: Livi ships with built-in evaluation tools and standard metrics as a baseline, and your team adds its own domain-specific metrics, scores, and thresholds on top — so every agent is measured against both our standard and yours.

Two ways, and you can mix them. Run your agent against a benchmark of test scenarios for automated, repeatable scoring, or test it manually — chatting with the agent yourself, reviewing individual responses, and scoring them by hand. Both feed into the same evaluation view, so you get a complete picture of how the agent performs across automated and human judgment.

Because you deploy with evidence, not assumptions. By the time an agent goes live, it has been measured against the metrics and thresholds you defined and tested against your benchmark and/or your own manual review — producing an overall agent score plus a breakdown of which checks passed and which need attention. Grounded in your uploaded content and governed by your guardrails (source citations, no self-diagnosis, auto-escalation), the agent ships validated to your standard. Our evaluation approach is grounded in peer-reviewed research on measuring the accuracy, safety, and trustworthiness of healthcare conversational AI — published in Nature Digital Medicine, Smart Health (ScienceDirect), IEEE, and arXiv.

It's interactive and iterative. As you build and refine an agent, you can evaluate it, see where it falls short, update your metrics, thresholds, or guardrails, and re-test — all in the same loop. Evaluation isn't a checkpoint you pass once; it evolves alongside your agent so the rules keep pace with how you want it to behave.

Livi is built for healthcare organizations and clinical and translational scientists — care teams, research groups, and digital health programs that want to turn their own approved content and real-world data into interactive agents. It's designed for teams that need engagement to reflect their clinical standards and stay under their control, whether they're running a study, supporting patients between visits, or scaling a care program.

No. Your uploaded guidelines, educational materials, and patient data are used to power and personalize your agent — not to train third-party foundation models or any model shared across customers. Your knowledge base stays yours.

Yes — Livi is built to be HIPAA-compliant from the ground up. Protected health information (PHI) is safeguarded throughout the platform, and PHI redaction is one of the default guardrails you can enforce on every agent. Patient data drawn from EHRs, wearables, and EMAs is used only to ground and personalize the agent's responses for that patient — never exposed beyond the controls you configure.

Livi connects to the sources you authorize: EHR records, wearable and biosignal data, and EMAs (delivered through our Centralive integration). Access is governed by your role-based permissions, so clinicians, researchers, and patients each see only what they should.

You do. Your approved clinical content, your real-world data, the agents you build, and the conversations they produce remain your intellectual property — Livi does not claim ownership of them and never reuses them for another customer. Livi retains the IP for the underlying platform and tooling (built on the open-source, peer-reviewed openCHA framework), but everything you create on top of it is yours to keep, export, and control.

Safety is enforced by configurable guardrails rather than left to chance. You can block diagnosis claims, prevent self-diagnosis, require every answer to cite its source, and set automatic escalation to a human provider when a conversation crosses a clinical threshold (for example, a patient reporting severe symptoms). These rules are encoded once and applied to every interaction.

Data is encrypted in transit and at rest, access is restricted through role-based permissions, and activity is logged for auditability. Security is layered: Livi maintains its own internal security standards, metrics, and technical safeguards as a baseline — and your team adds its domain-specific security and compliance requirements on top, so every agent meets both our standard and yours. Because guardrails like PHI protection and required source citations run at the response level, safety isn't only a network concern.