Use Cases
Scenarios that define the Rossoctl vision across verticals. Each maps to a persona from PERSONAS_AND_ROLES.md. Detailed drill-downs are linked where they exist.
For the taxonomy of agent operational models these scenarios target (read-only insight, synchronous task, async task, monitoring, event-driven), see Use Case Types.
Index
| Area | Personas | Drill-Down |
|---|---|---|
| Deployment & Access Control | Agent Developer | — |
| Platform Governance | Platform Operator | — |
| Identity & Trust | Platform Operator, Security Specialist | — |
| Observability | Platform Operator, Agent Developer | — |
| Agent Lifecycle | Platform Operator | — |
| Agent Discovery | Platform Operator | — |
| Multi-Tenancy | Platform Operator | — |
| Developer Experience | Agent Developer | — |
| Inbox Pattern | Agent Developer, Platform Operator, End User | — |
| Agentic Runtime & Sandbox | Agent Developer, Platform Operator | agentic-runtime/use-cases.md |
Deployment & Access Control
As an Agent Developer, I want to deploy an agent with access to both internal resources (database, memory service) and external resources (GitHub, S3, allow-listed domains), so that my agent can operate across trust boundaries without exposing internal services.
As an Agent Developer, I want to restrict access to my deployed agent to only myself, so that I can iterate privately before exposing it.
As an Agent Developer, I want to grant my team access to my deployed agent without requiring platform-level intervention.
As an Agent Developer, I want to grant other agents I have deployed access to my agent, so that I can compose multi-agent workflows via A2A without opening access beyond my own workloads.
Platform Governance
As a Platform Operator, I want to configure which inference providers and models are available per team, so that I can control cost and enforce approved model lists.
As a Platform Operator, I want to control egress to specific external endpoints per namespace, so that network policy is centrally managed.
As a Platform Operator, I want to apply a common guardrails policy to all incoming requests, so that content safety and compliance rules are enforced uniformly across agents.
Identity & Trust
As a Platform Operator, I want to integrate my build system with AgentCard resources to attach build provenance signatures, so that deployed agents carry verifiable evidence of their build origin and integrity.
As a Security and Identity Specialist, I want to sign an agent's A2A AgentCard with the workload's SPIFFE SVID, so that consumers can verify the agent's identity is attested by SPIRE before trusting its advertised capabilities.
Observability
As a Platform Operator, I want each namespace to route agent telemetry to its own observability backend (metrics sink, MLflow instance), so that teams have isolated visibility into their agents without cross-tenant data leakage.
As an Agent Developer, I want my A2A agent to be auto-instrumented with OpenTelemetry GenAI semantic conventions, so that traces, token usage, and tool calls are captured without manual instrumentation.
Agent Lifecycle
As a Platform Operator, I want to enforce that agents are admitted to a namespace only after evaluation at a restricted access level, so that untested or unverified agents cannot access production resources before demonstrating expected behavior.
Agent Discovery
As a Platform Operator, I want to list all running agents across namespaces and filter them by capability, skill, protocol, and other AgentCard attributes, so that I have a unified view of the agent fleet and can reason about platform-wide capacity and coverage.
Multi-Tenancy
As a Platform Operator, I want to define token usage quotas per role and namespace, so that I can enforce cost boundaries and prevent any single team or workload from consuming disproportionate inference capacity.
Developer Experience
As an Agent Developer, I want to quickly deploy a locally-developed agent into a remote cluster with scoped personal access, so that I can test against real infrastructure without a heavyweight CI/CD cycle.
As an Agent Developer, I want to deploy and test multi-agent workflows in a live namespace, so that I can validate agent-to-agent interactions under realistic conditions.
Inbox Pattern
As an Agent Developer, I want to define my agent's inbox source declaratively (queue topic, webhook path, event filter), so that the platform handles delivery and I focus on task logic.
As a Platform Operator, I want inbox-bound messages to carry authenticated sender identity through AuthBridge, so that authorization policies apply to asynchronous work items the same way they do for synchronous A2A calls.
As a Platform Operator, I want inbox-based agents to autoscale based on queue depth, so that compute matches demand without manual intervention.
As an End User, I want to submit a task to an agent and receive a handle I can poll for results, so that I am not blocked on long-running work.
Agentic Runtime & Sandbox
Detailed use cases for the sandboxed agentic runtime: coding agents, DevOps agents, tool-calling agents, MCP tool approval models, and sandboxing profile selection.
See: agentic-runtime/use-cases.md (PR #1005)