The Pavri platform works by identifying agents and assistants, attaching operational context, evaluating concrete actions against policy, preserving evidence, and triggering response workflows after detections or decisions.
Pavri maps identity, context, action, decision, evidence, and response into one accountable operating model.
Every major product surface is organized around the accountable action chain, not around disconnected logs or prompt-only controls.
Identity
Agent or assistant, owner, user, application, runtime, endpoint, environment, configuration, model, tools, and MCP relationships.
Decision
Matched policy and result. Framework and endpoint control points can allow, alert, or deny; the Local MCP Broker can hold configured MCP calls.
Evidence
Framework events, session records, endpoint observations, process events, configuration changes, decisions, and outcomes.
Text equivalent: The chain starts with identity, adds context, observes a concrete action, evaluates a decision, preserves evidence, and supports response.
Product surfaces
Discovery, action control, and investigation work as one system.
Inventory and identity inform policy. Policy decisions produce decision records. Evidence makes alerts and response actions investigable.
Discover
Find instrumented framework agents and enrolled coding-agent endpoints, including owners, tools, MCP relationships, workspaces, and recent activity.
Inventory
Control actions
Evaluate concrete operations at the control points available to each environment.
Allow / alert / deny
Investigate and respond
Move from alert to session, action, policy decision, evidence source, and response record.
Evidence chain
Next step
Walk through the Pavri platform.
Review the two launch environments and the evidence each contributes.
