Use Cases#
This page groups common Admin Console tasks and points to the pages that support each workflow.
Monitor platform health#
Use Home for high-level monitoring of request volume, reliability, cost, guardrail activity, and cache efficiency.
Common scenarios include:
- Daily health check — Review Total requests, Success rate, and Avg latency.
- Incident triage — Use a short time range and inspect Model reliability hotspots before opening Traces.
- FinOps review — Review Total spend, Cost breakdown, and model or user breakdowns.
- Governance monitoring — Review Top guardrail violations and success rate after policy changes.
- Optimization monitoring — Review Cache efficiency together with spend and token metrics.
For dashboard configuration, see Home. Allow time for new traffic to appear after you change providers, routing, or policies.
Analyze AI spend and usage#
Use Analytics → Usage and Cost when you need to move from organization-wide trends to a specific cost or usage driver.
Typical workflows:
- Monitor adoption — Review request volume and token consumption over time.
- Find cost drivers — Compare model and user spend.
- Investigate unusual activity — Look for sudden changes in requests, tokens, spend, or failures.
- Analyze a model, user, agent, or team — Open the corresponding activity tab to review scoped usage, cost, reliability, and model usage.
- Support budgeting or chargeback — Use cost and activity views to attribute spend to models, users, agents, or teams.
For page details, see Usage and Cost.
Measure optimization savings#
Use Analytics → Optimizations to measure AI cost savings from cache reads, prompt and context compression, memory retrieval, and related optimization features.
Typical scenarios include:
- Compare Actual and Would-have-paid cost.
- Identify whether cache reads, prompt compression, or memory retrieval contribute most to savings.
- Compare savings across models or teams.
- Validate savings after enabling context compression or other optimization features.
- Use Projected annual savings for planning and reporting.
For metric definitions and page layout, see Optimizations.
Apply policies and guardrails#
Use Config → Policies & Guardrails to apply safety, compliance, and content controls to Chat and API traffic.
| Use case | Typical guardrails | Policy pattern |
|---|---|---|
| Redact PII in user prompts | PII detection and redaction on Input, action Redact | Match the required request type; use Input stage; use 100% sampling for compliance-sensitive traffic |
| Block credential leaks | Secrets on Input and/or Output | Scope by organization or team; block or redact matched content |
| Harden against prompt abuse | Prompt injection, Jailbreak, Invisible text on Input | Apply to exposed Chat or API traffic and tune thresholds from observed violations |
| Control response safety | Toxicity, Code injection, Secrets on Output | Use Output stage to inspect model responses before return |
| Layered governance | Multiple guardrails linked to one policy | Use shared Conditions while each guardrail keeps its own type, threshold, and action |
Reuse guardrails across policies when the same control applies to different scopes. For a secrets walkthrough, see Block secrets.
For configuration, see Policies and Guardrails. For condition recipes, see Policy condition examples.
Integrate external systems with MCP#
Use an MCP server when Chat or agent workflows need controlled access to external tools or data sources.
Typical scenarios include:
- Access internal documentation or repositories.
- Retrieve data from external systems.
- Create or update records through supported tools.
- Call internal services or APIs.
For configuration, see MCP Servers. For a filled-in HTTP configuration, see Example: HTTP MCP server.
Investigate request issues#
Use Analytics → Traces when aggregate dashboards or usage charts are not enough to explain a failed request, latency spike, unexpected cost, routing result, or payload-level issue.
Typical scenarios include:
- Inspect the provider, model, and routing information for a specific request.
- Review request-level latency, token usage, and cost.
- Examine captured request and response data or request parameters.
For request-level investigation, see Traces.
See also#
- Home — Monitor organization-wide metrics and trends.
- Usage and Cost — Analyze spend, usage, and reliability.
- Optimizations — Measure AI cost savings.
- Policies & Guardrails — Apply governance controls.
- MCP Servers — Configure MCP integrations.
- Traces — Investigate individual requests.