Autonomous,
not uncontrolled.

Agents that act on your behalf need strong guardrails. PV is designed so that every agent is isolated, every action is recorded, and every permission is explicit. Security is built into the platform rather than added on top.

Each agent runs in its own secure sandbox, separated from other agents and other customers. Tasks cannot read each other’s memory, files, or credentials, so there is no data leakage between workloads.

• In transit: all traffic to and from PV is encrypted with TLS.
• At rest: agent memory, task data, and stored files are encrypted.

Agents follow the principle of least privilege. You decide which tools, integrations, and data each agent can access, and an agent cannot act outside the scope you grant it.

Every action an agent takes is logged and inspectable, including what it did, when, and with which permissions. You can review the full history of any task to understand and verify agent behaviour.

Agents run on compute nodes distributed globally for redundancy and low latency, so a failure in one region doesn’t take your workloads down with it.

You own your data. We process it only to provide the Services. See our Privacy Policy for details on collection, retention, and your rights.

If you believe you have found a security vulnerability in PV, please report it to us through our contact pagewith the topic “Support” and include enough detail for us to reproduce the issue.

Please give us reasonable time to investigate and fix the issue before disclosing it publicly, and avoid accessing or modifying other users’ data. We appreciate the work of security researchers and will acknowledge valid reports.