Skip to content

Migration guidesFrom Portkey

From Portkey, after the Palo Alto Networks acquisition

Portkey is a BYOK gateway: your provider keys live in its Integrations, and configs route, retry, fall back and cache. Since May 2026 it is part of Palo Alto Networks as the Prisma AIRS AI Gateway. Most of its routing maps directly onto INFRO routes; the prompt tooling does not.

Acquired by Palo Alto Networks (closed 2026-05-29); now Prisma AIRS AI Gateway

INFRO accounts are self-serve. The catalog, the docs, and the API contract are public, the gateway is deployed, and anything beyond v1 is marked Later. Plans and trial

Where things stand

The acquisition

Palo Alto Networks completed its acquisition of Portkey on 2026-05-29; Portkey becomes the core AI gateway for Prisma AIRS. Portkey's pricing page now says it is Prisma AIRS AI Gateway and still lists its Developer, Production, Enterprise and Open Source tiers.

The open-source gateway

Portkey-AI/gateway is MIT and not archived. Its last release is v1.15.2 (2026-01-12) and its last push was 2026-05-25. We found no statement from either company about its future.

Whose account serves the request

Yours. Provider keys sit in Portkey Integrations, and your providers bill you, as with INFRO.

Nothing about the acquisition forces a move; Portkey still runs. Stay if one of these is why you use it.

  • You are buying into Palo Alto Networks' security stack, and the gateway is part of that purchase.
  • You use the Prompt Engineering Studio or prompt library. INFRO has no prompt management.
  • You use semantic caching, or conditional routing on arbitrary request metadata. INFRO caches exact matches and its route rules read a fixed set of request fields.
  • You self-host the open-source gateway and want to keep doing that without an agreement.
  • You use Portkey for batches or a provider INFRO does not connect to.

The edits, one row each. Everything not listed here stays as it is.

 PortkeyINFRO
Base URLhttps://api.portkey.ai/v1 (PORTKEY_GATEWAY_URL).https://api.infro.io/v1
KeyA Portkey API key, as x-portkey-api-key or the SDK's apiKey.An INFRO gateway key in Authorization: Bearer. Provider keys move from Integrations to INFRO connections.
Model id@openai-prod/gpt-5.6-terra, or x-portkey-provider plus a bare id.openai/gpt-5.6-terra. To pin a provider, routing.providers.allow; for a named setup, a route.
Headersx-portkey-config, x-portkey-metadata, x-portkey-trace-id, x-portkey-cache-namespace.A route as model; metadata in the body (or X-INFRO-User-Id, X-INFRO-Feature); X-INFRO-Trace-Id; X-INFRO-Cache-Namespace.

Before · Portkey

import OpenAI from "openai";
import { PORTKEY_GATEWAY_URL } from "portkey-ai";

const client = new OpenAI({
  apiKey: process.env.PORTKEY_API_KEY,
  baseURL: PORTKEY_GATEWAY_URL,
  defaultHeaders: {
    "x-portkey-config": "pc-support-fallback",
    "x-portkey-metadata": JSON.stringify({ _user: "user_123" }),
  },
});

await client.chat.completions.create({
  model: "@openai-prod/gpt-5.6-terra",
  messages: [{ role: "user", content: "Hello" }],
});

After · INFRO

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.INFRO_API_KEY,
  baseURL: "https://api.infro.io/v1",
  defaultHeaders: { "X-INFRO-User-Id": "user_123" },
});

await client.chat.completions.create({
  // A route holding the config's targets and failure rules
  model: "support-fallback",
  messages: [{ role: "user", content: "Hello" }],
});

Configs become routes

strategy.mode: fallback is a priority route, loadbalance a weighted route, retry the route's failure rules. Conditional routing becomes route if-rules where the condition reads a field INFRO's rules support.

Integrations become connections

Each Integration's provider key becomes a connection. Several keys of one provider are spread across, with standby keys used last.

Budgets move up a level

Portkey sets budgets on an Integration per workspace. INFRO sets them on the organization, a project, a member, a key or an end customer, and refuses at the ceiling before the call.

Portkey's objects against INFRO's, from Portkey's public documentation read on 19 October 2026 and checked against what the gateway implements. Two rows say INFRO has nothing.

PortkeyWhat it does thereIn INFRO
Integrations and AI Providerssame shapeA credential vault plus provider slugs with models, budgets and limits.Provider connections, sealed, several keys per provider, models per key.
Configs: fallback, loadbalance, retrysame shapeJSON routing objects, saved as pc-... and attached to keys.Routes with priority or weighted targets and failure rules per kind; keys can be pinned to routes.
Conditional routingdifferentRoute on metadata or parameters.Route if-rules on a set of request fields, plus a prompt classifier. Not on arbitrary metadata.
Cache (simple and semantic)differentExact and semantic response caching.Exact-match only, opt-in, with a namespace.
Budget and rate limitsdifferentUSD or token limits per Integration per workspace, with alerts.Dollar budgets at five scopes, 402 before the call, alert thresholds; request-rate tiers per key. No token limits.
GuardrailsdifferentA catalogue of guardrails in the gateway.Pattern rules that block, redact or flag, and checks on a decision model.
Logs, traces, feedbackdifferentRequest logs, traces and user feedback.Request log and traces with estimated cost, OpenTelemetry export. No feedback scores.
Prompt library and studionothingPrompts managed and versioned outside the code.Nothing. Prompts stay in your code.
MCP gatewaysame shapeMCP servers behind the gateway.An MCP gateway for your own MCP servers, with approved tools and grants per key.
Workspaces, RBAC, SSOsame shapeWorkspaces and roles; SSO on enterprise.Organizations and projects, roles, project access per member; SSO and SCIM from the Business plan.
BatchesnothingBatch requests through the gateway.Nothing.
Self-hostingdifferentMIT open-source gateway; enterprise hybrid.Hosted. Self-hosted or hybrid is an enterprise agreement.Enterprise — Available on an enterprise agreement rather than self-serve.

Our reading of Portkey’s published documentation, not a trial or a benchmark. If we have read something wrong, tell us and we will fix the row.

What INFRO does not do

  • Prompt management, feedback scores or evaluations.
  • Semantic caching, or routing on arbitrary metadata.
  • Batches, embeddings, files or fine-tuning.
  • Token-count limits; INFRO limits by request rate and by dollars.
  • Run on your infrastructure without an enterprise agreement.

What does not come across

  • Logs and traces in Portkey. There is no importer; export what you need.
  • Saved prompts and their versions.
  • Configs, budgets and guardrails as objects. Recreate them as routes, budgets and guardrails.

Questions

Does the acquisition mean Portkey is shutting down?
Not on anything we read. Palo Alto Networks completed the acquisition on 2026-05-29 and Portkey's pricing page still lists its tiers. We found no statement about the open-source gateway's future.
How do Portkey configs map to INFRO?
Fallback configs become priority routes, load-balance configs weighted routes, retries the route's failure rules. Send the route's name as model.
Who pays the model providers?
You do, directly, as with Portkey. INFRO is a subscription and adds 0% to model spend.

Move with us, or on your own

The migration program maps your Portkey setup to INFRO connections, routes and budgets on a call and moves the first service with you. Your providers keep billing you directly; INFRO is a subscription and adds 0% to model spend.

Start 14-day trial Book a migration sessionINFRO vs PortkeyRead the quickstart

Sources

Every statement on this page about Portkey comes from one of these, read on 19 October 2026. Nothing is inferred from a trial, a benchmark, or a conversation.