Skip to content

Migration program

Move your AI traffic to INFRO. We run the switch with you.

For teams moving from Helicone, Portkey or a self-run LiteLLM proxy. You keep your provider accounts and contracts; INFRO adds routing, failover, budgets and cost per request in front of them.

Start 14-day trial Book a migration session

A guided migration, not a sign-up form.

Migration session

We map your setup and move the first service with you

On a call, we map your keys, configs, fallbacks and tags to INFRO connections, routes and budgets, then move one service together before you move the rest.

Direct channel

A line to the INFRO team

A direct channel during the migration and after it, for questions, issues and feature requests, instead of a ticket queue.

Your accounts stay yours

Nothing about your provider contracts changes

You connect the provider accounts you already hold and keep paying your providers directly at your own rates. INFRO takes 0% of model spend.

The short version. Each card links to the full guide or comparison, including the cases where staying is the better answer.

From Helicone

Acquired by Mintlify; in maintenance mode

  • Base URL https://api.infro.io/v1 and an INFRO key; Helicone-Auth comes out.
  • Helicone-User-Id and Helicone-Session-Id become metadata.user_id and metadata.session_id; Helicone-Property-* becomes metadata.tags.
  • Request history, prompts and evaluator scores don't carry over. Export them first.
Read the Helicone guide

From Portkey

Now Palo Alto Networks' Prisma AIRS AI Gateway

  • Virtual keys become provider connections, plus scoped gateway keys with model allowlists, IP ranges and rotation.
  • Configs with fallback or load-balance targets become routes, with per-failure rules for 429s, 5xx and timeouts.
  • Budgets are on every plan. The prompt library and log history don't carry over.
How INFRO and Portkey compare

From LiteLLM

Self-run proxy or in-process library

  • Keep the library and point it at INFRO as an OpenAI-compatible upstream, or retire the proxy.
  • model_list entries become provider connections; router fallbacks become routes.
  • No Postgres, Redis, upgrades or price tables to run. Self-hosting needs an enterprise agreement.
How INFRO and LiteLLM compare

Most teams change a base URL, a key and the model name. The work is in mapping what you had.

  1. 1Connect your provider accountsThe keys you already hold, sealed in INFRO.
  2. 2Write routes and budgetsFailover order, standby keys, and limits that apply before the provider call.
  3. 3Move one serviceChange the base URL and key, then compare INFRO's request log with the one you had.
  4. 4Move the restService by service, at your pace.
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.infro.io/v1",
  apiKey: process.env.INFRO_API_KEY,
});

await client.chat.completions.create({
  model: "anthropic/claude-sonnet-5.5",
  messages: [{ role: "user", content: "Hello" }],
  metadata: { user_id: "cust_123", feature: "summaries" },
});

0%
markup on model spend
$35–$599
a month, flat, by plan
14 days
trial of any plan, card at checkout
Off
prompt recording, until you switch it on

Launch offer: 50% off your first two months on any monthly plan with the code PH50, until 18 November 2026. It applies to the INFRO subscription; your providers keep billing you directly for model use.

How budgets hold under load: a $1 cap at 500 concurrent requests, method and results.

Book your migration session

Reply to our message, or write to hello@infro.io with what you run today. We’ll schedule the session and come to it with a map of your current setup. If the honest answer is that you should stay where you are, that is the answer you will get.

Start 14-day trial Contact usRead the quickstart