One platform, every frontier model.
No single model is best at everything. Kroolo Work stays model-agnostic, so you are never tied to one vendor.
Match the model to the use case.
Three tiers, each suited to a different kind of work. Kroolo routes between them automatically.
Tier placement is illustrative. Available models change as providers release new versions.
Every request goes to the right model.
Routing runs in the background. Your teams just ask; Kroolo decides how to answer.
One call. The right model.
Send the same request from any language. Kroolo picks the model, applies your team policy and records the answer for audit.
import Kroolo from "@kroolo/sdk";const kroolo = new Kroolo({ apiKey: process.env.KROOLO_KEY });const reply = await kroolo.chat.create({model: "auto", // Kroolo picks the right modelteam: "finance", // team policy and spend cap applymessages: [{ role: "user", content: "Summarise this week's risks" }],});
{"model": "claude-sonnet","tier": "balanced", // chosen for this task"cost_usd": 0.0021,"audit_id": "aud_8f21c"}
Frontier quality, without frontier bills.
Smart routing is how Kroolo keeps AI affordable as usage grows across the organisation.
Illustrative only. Actual mix depends on your workflows.
You decide who uses what.
Admins choose which models are available in the workspace, set who can use each one by team, and cap usage so spend stays predictable.
- Turn models on or off for the whole workspace
- Control access by team or department
- Set spend limits per team
See where every token goes.
One place to see how AI is used across the organisation: which models, which teams, and what it costs. Routing keeps routine work on efficient models.
- Usage by model, team and project
- Allocate AI cost back to the team that used it
- Alerts before a team reaches its cap
Bring your own model or key.
Use the providers you already contract with, or connect a model you host yourself. Kroolo routes across them with the same controls.
- Connect your own provider keys
- Use a self-hosted or private model
- Same access, spend and audit controls on every model
Every model, same guardrails.
Whichever model answers, the same permission checks, guardrail policies and audit trail apply.
FAQs on Model Hub
See smart routing in action.
Bring a real workflow. We will show which model handles each step and what it costs.
More from Kroolo AI
- Responsible AI Framework for AI AgentsKroolo's responsible AI framework for AI agents: nine principles, five autonomy risk tiers, lifecycle controls, customer commitments and standards mapping.
- AI Agents for WorkUse 50+ prebuilt AI agents for projects, sales, marketing, HR and data, or build your own. Payments and external sends wait for your approval.
- Agentic AI Development Services, Idea to ProductionKroolo's AI engineering team turns your problem statement into a governed, production AI product in 7 to 18 weeks, deployed in your cloud and run after launch.
- Enterprise AI Guardrails and Sovereign AI18 guardrail policies on every AI request, human approvals, a full audit log and deployment in your AWS, Azure, Google Cloud or data centre. See how it works.
- LLM Routing and AI Cost Control by TeamKroolo routes each AI request to a fast, balanced or frontier model, fails over when one is slow, and shows AI spend by team, workflow and agent. See it live.
- Enterprise AI StackSee how Kroolo's enterprise AI stack works: connectors, a knowledge graph with vector and keyword search, permission checks, agents and cited answers.