Feature × Audience

How do MSPs route model requests with Plugsky?

MSPs route with Plugsky by giving each client workspace its own strategy: a cheap-first default handles routine volume, custom rules escalate hard tickets, and premium strategies are reserved for clients who pay for them. Routing stays inside the shared platform, keys remain per client, and flat pricing means more resolutions do not erode margin.

Key facts

Router modelmodel="plugsky-fusion" applies a strategy per client workspace (live)
ScopingStrategies and rules per workspace or API key, one project per client
Defaultscost_saver or balanced for routine volume; max_quality for premium queues
Cost modelFlat self-serve plans; no per-token charges, fair-use RPM per tier
White-labelDeliver routing-driven services under your own brand
ObservabilityModel, strategy and rule logged per request; tokens returned in responses
ResidencyPin each client workspace to the region their contract requires
RoadmapClassifier routing /v1/plugsky/route (model=auto) is coming soon

TL;DR

  • Template one routing policy and stamp it across client workspaces.
  • Keep routine traffic on cheap-first strategies and escalate deliberately.
  • Reserve premium strategies for clients who pay for premium outcomes.
  • Use per-client logs for margin reviews and client reporting.
  • Pin each workspace to its contracted region without changing code.

How it works, step by step

  1. Design a small set of strategy templates: economy, standard and premium support.
  2. Create each client workspace with its template attached and keys scoped to that project.
  3. Add escalation rules only for markers you trust — ticket complexity, language or tool use.
  4. Track routing distribution per client monthly to spot templates that cost more than they should.
  5. Publish routing summaries in your portal to show where quality investment went.
  6. Adjust one template at a time and roll it out to matching clients.
1Design a small setof strategytemplates: economy,2Create each clientworkspace with itstemplate attached3Add escalationrules only formarkers you trust —4Track routingdistribution perclient monthly to5Publish routingsummaries in yourportal to show6Adjust one templateat a time and rollit out to matching

Try it yourself

Open the AI workload router simulator →

Strategy templates per client tier

MSPs thrive on repeatability. Instead of tuning models per client, define two or three routing templates — economy, standard, premium — and attach one to each client workspace. Economy runs cost_saver with escalation on genuinely hard prompts; standard runs balanced; premium runs max_quality for the queues that justify it.

Because strategies can be scoped per workspace or API key, a change to a template propagates through a review-and-apply process, not a code deploy. That is the operational difference between managing a platform and firefighting a stack.

Escalation that protects margin

Margin is decided at the escalation boundary. If every ticket can reach a frontier model, premium cost leaks into economy contracts. Write rules with clear markers: tickets that contain error codes or tool calls escalate; FAQs and status checks stay cheap.

  • Measure: the decision log shows escalation rates per client, so a mis-tuned threshold is visible before it becomes a margin problem.
  • Cap: where contracts require it, pin a fixed model for a client to bound outcomes and cost.
  • Flat pricing: no per-token charges means more traffic does not mechanically shrink gross margin.

Client reporting without leaking your economics

Clients want evidence their service is run well. Routing logs give you the story in neutral terms: most requests answered instantly on efficient models, complex cases escalated for better outcomes, and residency honoured. Token telemetry returns in every response for your internal reviews, even though it does not drive billing.

Keep the platform white-label so client-facing material carries your brand, and keep region pinning per workspace so a client expanding into a new jurisdiction gets the right deployment without a re-platforming project.

Honest comparison

ConcernPlugsky routingPer-client model stacksReselling per-token APIs
Per-client policyStrategy templates per workspaceDuplicated configurationAccount-level settings
Escalation controlOrdered rules with clear markersManualLimited
Margin visibilityRouting logs and token telemetryCustom reportingProvider invoices
White-labelFull brand controlFullProvider branding
Ops loadTemplate changesPer-client maintenanceBilling reconciliation

Frequently asked questions

How do we give each client a different policy?

Create one workspace per client and attach a strategy template — economy, standard or premium. Strategies can be scoped per workspace or key.

How do we stop premium models leaking into cheap contracts?

Escalate only on reliable markers and review the decision log monthly. If a client must be bounded, pin a fixed model for their keys.

Does routing raise our costs as clients grow?

Self-serve plans are flat with unlimited fair-use usage and no per-token charges, so growth changes request rate, not per-token spend.

Can each client have its own region?

Yes. Pin each workspace to the contracted region, or use VPC, on-prem and air-gapped deployment for stricter contracts.

Can we white-label the routing layer?

Yes. The routing behaviour is internal to your delivery; clients see your portal and your brand.

What do we show clients about routing?

Use the decision log for internal reviews and share neutral summaries — efficient handling, escalation on hard cases, residency compliance.

How quickly can we onboard a client?

Attach a template, create scoped keys, import their content and point their channel at your service — typically the same day.