Your data remains yours.
You retain ownership of your business data, task inputs, and outputs. Define which context, results, and audit records the cloud platform may process, with agreed retention and residency.
Read the designAI belongs inside your business’s standards. Our platform principles put explicit authority, verifiable work, and human accountability at the center. Your data remains yours, with the option to run agent VMs and connections inside your own VPC.
You retain ownership of your business data, task inputs, and outputs. Define which context, results, and audit records the cloud platform may process, with agreed retention and residency.
Read the designRun your own agent VMs and host your own connections in your network. Keep source systems and credentials in your environment, and choose your model endpoints and permitted data flows.
Read the designA Principal grants an agent a limited subset of their authority. A session runs with an execution identity that specifies who it acts for, what it can access, and for how long.
Read the designPolicies define permitted actions, approval requirements, time, and cost limits. Enforcement belongs in the runtime and connected systems, outside the language model.
Read the designPersonal sessions are private by default in the design. Reading, continuing, forking, and collaborating require deliberate sharing decisions and appropriate access.
Read the designEvidence, structural checks, and acceptance criteria make results reviewable. Record tool activity, approvals, policy decisions, and outcomes according to an agreed retention policy.
Read the designConnections supply authenticated access without exposing credentials directly to the agent. Use least privilege, secret references, rotation, and source-system access controls.
Read the designAgent, Task, Platform, and Capability Owners have distinct accountabilities. Quality and cost need ongoing measurement, with failures routed to the responsible owner.
Read the design