Core Services#

Core Services in the Admin UI sidebar groups AI Gateway, access control, governance, and FinOps pages.

Use this section when you need to connect providers, onboard users, enforce policies, attach MCP tools, or monitor cost and reliability.

Next: if the organization is new, start with First Steps. Otherwise follow Setup → Govern → Monitor and optimize, or jump from the menu section overview.

Setup#

Providers — Connect vendor endpoints, enable models, and set per-provider limits. Configure Routing Rules and see Routing best practices when you need failover or traffic splitting.
AI Access — Invite users, set spend and quota limits, choose allowed providers, and enable context compression.
MCP Servers — Register optional Model Context Protocol integrations so Chat and agents can call external tools.

Policies & Guardrails — Enforce safety, compliance, and content controls on Chat and API traffic. Configure Guardrails and Policies on separate tabs in the same Admin area.

Home → Dashboards — Check cost, volume, reliability, guardrail, and cache trends at a glance.
Optimizations — Compare would-have-paid vs actual spend and measure savings by technique.
Analytics → Usage and Cost — Drill into consumption, spend, and activity by model, user, team, or agent.
Analytics → Traces — Inspect individual requests for latency, routing, tokens, cost, and payloads.

Table 1: OptScale AI Core Services — sidebar menu items
Menu item	Purpose
Home / Dashboards	At-a-glance operational health for cost, volume, reliability, and cache efficiency.
AI Access	Onboard members; manage limits, teams, allowed providers, and compression.
Providers	Connect AI vendors, enable models, and set quotas.
Routing Rules	Match conditions, weighted targets, and fallbacks for Chat and API traffic.
Routing Best Practices	Failover, weighted split, Chat request types, and other routing patterns.
Policies & Guardrails	Overview—policy vs guardrail, evaluation workflow, and setup order.
Guardrails	Reusable checks (PII, secrets, toxicity)—create, edit, and troubleshoot guardrail definitions.
Policies	Enforcement scope—conditions, sampling, timeout, linked guardrails, and usage statistics.
MCP Servers	Register MCP servers and control which tools are enabled or auto-run.
Optimizations	Measure projected and realized AI cost savings from optimization features.
Usage and Cost	Under Analytics — FinOps views for tokens, spend, and scoped activity.
Traces	Under Analytics — Per-request audit and debugging for AI traffic.