Built for teams that are crushing tokens.

Concentrate is an AI gateway and model router for coding agents, internal workflows, and customer-facing products, with the governance that fintech, insurtech, and healthcare teams need, and any company that processes sensitive data.

Markup on tokens
None
Infrastructure you run
None
Cloud or framework lock-in
None

We supply the inference

Buy tokens for every model through Concentrate. No contract with each lab, and no commitment to any single one.

Nothing to host or patch

Concentrate is fully managed. There is no proxy to deploy, no database to run, and no upgrade to schedule.

Bulk pricing as you grow

We pool usage across customers and hold the provider commitments ourselves, so your rates improve as your spend grows.

How to use Concentrate

  1. STEP 01

    Create your workspace and add teams.

    Workspace
    WorkspacePersonal
    TeamSupport
  2. STEP 02

    Purchase tokens

    Use credits on any model or provider. No service fee or credit card fees.

    Balance$99.00
    Used today$10.42
  3. STEP 03

    Swap your base URL

    . Point your client at Concentrate. Pick the model you want.

    API key
    base_url = "https://api.concentrate.ai/v1"
  4. STEP 04

    Deploy and govern

    Add spend controls, review logs, and adjust as you grow.

    Controls
    Spend limit$500/key
    AccessTeam RBAC
    Audit logsOn

Frequently asked questions (FAQs)

What is Concentrate.ai?

Concentrate is a managed AI gateway and model router. One API key and one base URL reach every major model, including decision models like TypeSafe Jev. We supply the inference and manage provider capacity and failover. Your team sets which models each team can use, spend limits, and data rules like ZDR.

What is an AI gateway?

An AI gateway, also called an LLM gateway or LLM proxy, sits between your apps and the model providers. Your apps call one API. The gateway handles provider keys, failover, spend limits, logs, and access rules for every model. Concentrate runs the gateway for you, so there is nothing to deploy.

Is Concentrate a model router?

Yes. You choose the model, and Concentrate routes each request to a provider that serves it, following the fallback order you set.

Do I need accounts or keys with each provider?

No. You buy tokens for every model through Concentrate, so there is no contract or key to set up with each lab. If you already have provider credits or committed spend, add those keys with BYOK and use them through the same API.

How does pricing work?

You pay for the tokens you use, with no markup, no service fee, and no credit card fees. We pool usage across customers and hold the provider commitments ourselves, so your rates improve as your spend grows without you signing a commitment.

How can we lower model spend?

Our model optimizer, now in beta, finds the jobs your agents repeat and tests cheaper models, including open models like GLM, DeepSeek, Qwen, and Kimi, on your real requests. You see the cost and quality side by side, then choose whether to switch. The optimizer is not available for HIPAA organizations.

How is Concentrate different from OpenRouter?

OpenRouter charges 5.5% on every credit purchase, or 8% on its Business plan. Concentrate has no markup, no service fee, and no credit card fees, and pooled rates improve as your spend grows. You also get spend limits by team or key, ZDR, SSO, RBAC, audit logs, and one BAA for HIPAA workloads.

How is Concentrate different from Vercel AI Gateway?

Vercel charges list price and passes payment processing fees on to you unless you move to invoiced Enterprise billing. Concentrate has no markup, no service fee, and no credit card fees. It is also not tied to one cloud or framework: it works from any host with the OpenAI, Anthropic, or Vercel AI SDK, LiteLLM, or a coding agent. Our BAA covers the inference itself.

How is Concentrate different from LiteLLM?

LiteLLM does not include inference. You bring your own: sign a contract with every lab, set up keys, billing, and rate limits with each provider, and reconcile a separate invoice from each one. Then you host, patch, and upgrade the proxy yourself. Concentrate is fully managed and includes the inference with no markup, so one API key, one contract, and one invoice cover every major model.

How is Concentrate different from Portkey?

Portkey does not include inference either. You connect your own provider accounts, which means signing each lab's contract, managing each provider's keys, billing, and rate limits, and paying each invoice. Concentrate supplies the inference with no markup, so one API key and one contract cover every major model, with spend limits, logs, and access controls built in.

Is Concentrate built for regulated industries?

Yes. Fintech and insurtech teams use Concentrate, and so do healthcare companies and any company that processes sensitive data. They govern AI usage in one place: SOC 2 Type II, one BAA for HIPAA workloads, zero data retention, PII, PCI, and PHI redaction, SSO, RBAC, and audit logs across every provider.

Can we use Concentrate for HIPAA workloads?

Yes. Sign one BAA with Concentrate. In a HIPAA organization, requests only go to BAA-covered routes and zero data retention stays on. The model optimizer is not available for HIPAA organizations.

Use any model in under 30 seconds.

Create a key, point your client at Concentrate, and pick a model. Your existing OpenAI or Anthropic SDK code keeps working.

base_url = "https://api.concentrate.ai/v1"