One key for AI workflows

Models and AI services. One API. One balance.

Build through the API, or work in Primex Chat and Create. See your routing and usage costs in one place.

No credit card required to sign up.

Familiar SDKsPay as you goRoutes & costs

Models · AI services · one key

Example workflowExplore the possibilities

From invoice to usable data.

Inputinvoice_q3.pdf4 pages
One APIDocument Analysis (Invoice)
Structured fields
Vendor
Acme Corp
Invoice
INV-3391
Total
$4,820.00

Explore Primex

Your models. Your tools. One place.

Choose the intelligence for your work, then add the capabilities around it.

Routing

Routing that fits your workload.

Cost-focused partner routes or declared infrastructure. Two approaches for different workloads.

Choose declared infrastructure

Verified

For workloads where a declared infrastructure tier comes first.

Your requestVerified routeResult
  • Declared infrastructure tier
  • Served tier recorded in Usage
  • Request-level usage and cost

Know the infrastructure tier behind each request, alongside its cost.

Optimize for cost

EconomyComing soon

For experiments and everyday workloads where cost comes first.

Your requestPartner routesResult
  • Cost-focused partner routes
  • Flexible upstream selection
  • Fallback within the routing policy

Let the route adapt to the workload, with cost as the priority.

Today, requests use Auto routing. Check the Catalog for the infrastructure tiers available to each API.

Developer quickstart

Your tools. Connected to Primex.

Already using OpenAI or Anthropic?

Choose a compatible model, then update your base URL and API key. Keep building with the SDK you know.

Base URLapi.openai.comgateway.primexos.com
API keysk-...prx-...

Compatibility is endpoint- and offering-specific. The live Catalog and docs are authoritative.

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "prx-...",
  baseURL: "https://gateway.primexos.com/v1"
});

const response = await client.responses.create({
  model: "YOUR_MODEL_ID",
  input: "Build something useful."
});
  1. 01

    Get your API key

    Create an account and generate a prx- key.

  2. 02

    Try a request

    Run it in Playground or connect your existing SDK.

  3. 03

    Add credits when ready

    Top up your AI balance and pay for measured usage.

Why Primex

Know what happens to every request

See how each request was routed, what it used, how long it took, and what you were charged.

  • Request-level usage

    See the model or service, measured units, request time and final net cost, attributed to the key that sent the request.

  • Routing visibility

    Check the requested mode and, where recorded, the tier that served it. Routing stays visible alongside usage and cost.

  • Spending controls

    Failed requests aren’t billed. Set monthly spend caps for your application keys and track their spending in Usage. See the FAQ for how caps work.

EXAMPLE USAGEreq_7f3a2c9e
Status
Succeeded
Model
gpt-5.6-sol
Requested mode
Auto
Served tier
Verified
Measured usage
1,240 in · 318 out tkn
Request time
2.31 s
Net cost
$0.0043

Live models & pricing

Start with what you need. Pay for what you use.

A few useful starting points from the live Catalog. One AI balance covers models and services, each priced in the unit it measures.

View all models & pricing

Open the Catalog for current availability and pricing.

Frequently asked questions

Getting started, connecting your tools and understanding your costs.

How do I get started with Primex?

Create a Primex account, then try the Playground or open Chat and Create. For your own application, generate a prx- key and follow the client setup guide. Primex-managed APIs do not require separate upstream provider accounts.

Will my SDK or coding tool work with Primex?

Primex supports the OpenAI SDK on compatible chat and Responses endpoints, and the Anthropic SDK on compatible Claude Messages offerings. Setup guides cover Claude Code, Cursor, Windsurf, Continue and Cline, and MCP. Support depends on the selected model, endpoint and client features; check the Quickstart and live Catalog.

How do Economy and Verified routing differ?

Verified describes declared infrastructure; current requests use Auto routing on that tier. Economy is planned as a cost-focused option using partner routes, with flexible upstream selection and fallback within its policy. It is not available to select yet. See routing options for the comparison and the Catalog for current offering details.

How do credits, plans and prompt caching work?

Add prepaid AI credits and pay for measured usage. No subscription is required; optional plans have rolling allowances and supported-usage limits. The Catalog lists each offering’s billing unit and price, with cache read, cache write and long-context rates where published. Any welcome credit is subject to account eligibility and required verification. See Billing and Credits.

Are failed requests charged, and can I control spending?

Failed requests aren’t billed. Usage shows each request’s outcome, measured units and final net cost. Set an optional monthly spend cap for each customer API key. Caps apply to net recorded usage; a request already in flight can finish above the remaining amount.

How is my data handled?

Primex sends the inputs needed for your request to the upstream provider serving it and records usage and operational data for billing, support and reliability. Storage and retention differ across APIs, Chat and Create; upstream providers apply their own policies. Read the Privacy Policy and Policies.

What payment methods are supported?

Cryptocurrency payments are supported. Any additional payment methods available for your order are shown at checkout.

Start with Primex

Your next idea starts here.

Build with the API, think with Chat, or make something in Create. Start with one Primex account.

No credit card required to sign up. No subscription required.