Use Cases
Most teams do not need a different application architecture to use Olyx. They need a governed path for model calls, a trace they can inspect later, and a clear boundary for what leaves the customer environment.
Common Patterns
| Use case | Why Olyx helps | Public docs |
|---|---|---|
| Support copilots | Record each request, compare replays, and inspect routing behavior before changing production logic. | Examples |
| Routing by cost or latency | Use trace summaries and routing decisions to compare models before switching traffic. | Cost Intelligence |
| VPC and private deployments | Keep provider secrets customer-side and use customer-managed telemetry when the project boundary requires it. | Policy Schema |
| MCP tool flows | Track tool calls, parent/child steps, and replay outcomes for agentic workflows. | Minimal Protocol |
| Audited production AI | Apply guardrails before execution, then keep the trace and ledger as the audit record. | Governance |
Support Worker Flow
The CLI checks policy, your app makes the request locally, and Olyx records the trace and replay metadata.
export OLYX_API_KEY=ak_...
export OLYX_BASE_URL=https://app.olyxai.io
olyx inspect
olyx config
olyx trace --metadata '{"source":"support-worker","feature":"ticket-reply"}'
The model call still happens in your codebase or runtime. Olyx stays on the request path to enforce policy, classify the trace, and make the run visible later.
Private Deployment Pattern
For VPC deployments and customer-managed boundaries, the same high-level flow still applies. The difference is what the project policy allows to leave the environment.
{
"distribution": {
"telemetry_mode": "customer_managed",
"payload_storage_mode": "customer_managed",
"payload_retained": false,
"replay_execution_mode": "disabled",
"secrets_boundary": "customer_environment",
"cloud_payloads": false,
"local_ref_required": true
}
}
When this boundary is active, use an opaque local_ref to point back to your
own database row, object store key, or replay record. Keep provider secrets and
private payloads in the customer environment.
What To Read First
If you are deciding which docs to open next, start here:
- CLI for install and trace creation.
- Authentication for API keys and dashboard sessions.
- Protocol for the minimal control-plane flow.
- Governance for ledger review and weekly exports.
- Cost Intelligence for comparing models and replay results.